bookkeeping.n_examples.note: n_examples reconciliation: the screen samples a frozen 160-prompt subset of the 2000-item Task-A pool (960 completions per sampling seed per rendering); the smoke legs draw 6 prompts per step from the same frozen 2000-item pool for 100 steps. The experiments.json rows record 160 for screen legs and 2000 for smoke legs by that construction; the figures describe different stages of one frozen dataset, not an inconsistency. caveat.delta_comparability.statement_1: The base emission rate is measured on 1920 untouched-policy rollouts over 160 prompts and two sampling seeds; the final selection rate is measured on 36 completions of the final train batch (72 pooled over two training seeds), a factor of roughly 53 fewer, on a different prompt subset, sampled on-policy from a just-trained model. caveat.delta_comparability.statement_2: A 36-completion rate carries a Wilson 95% half-width of roughly 0.15, so every individual base-to-final delta in this study is smaller than its own sampling noise and no single delta is distinguishable from zero. caveat.delta_comparability.statement_3: Corrected strength of the delta finding: across four points at two scales every delta is small and three of four are negative, so this study finds NO EVIDENCE of learned seam selection beyond the base policy's pre-existing emission; at this power that is absence of evidence, not evidence of absence, and it does not support a claim that no learning occurred. descope.7b.bold_final_answer.executed_seeds: 3 descope.7b.bold_final_answer.halfwidth_cost_of_descope: 0 descope.7b.bold_final_answer.per_pre_stated_descope_order: 0 descope.7b.bold_final_answer.pre_registered_seeds: 3 descope.7b.bold_final_answer.wilson_halfwidth_at_executed_n: 0.055512631074 descope.7b.bold_final_answer.wilson_halfwidth_at_n108_hypothetical: 0.055512631074 direction.curve_seams_below_separator_lower_edge_7b: 5 direction.curve_seams_with_negative_differential_shift_7b: 4 direction.curve_seams_with_positive_differential_shift_7b: 1 direction.slate_mean_differential_shift_7b: -0.19125 equivalence.3b.boxed.diff_mb6_minus_mb12: -0.0277777777778 equivalence.3b.boxed.equivalence_pass: 1 equivalence.3b.boxed.fisher_p: 1 equivalence.3b.boxed.micro12_rate: 0.444444444444 equivalence.3b.boxed.micro12_wilson_hi: 0.604189382443 equivalence.3b.boxed.micro12_wilson_lo: 0.295412687401 equivalence.3b.boxed.micro6_inside_micro12_wilson: 1 equivalence.3b.boxed.micro6_rate: 0.416666666667 equivalence.3b.boxed.newcombe_hi: 0.191853953331 equivalence.3b.boxed.newcombe_lo: -0.243692614566 equivalence.3b.boxed.same_side_of_gate: 1 headroom.1p5b.backtick_hash_span.exploitation_given_emission: 0.466666666667 headroom.1p5b.bold_final_answer.exploitation_given_emission: 0.245557350565 headroom.1p5b.boxed_numeric_final_line.exploitation_given_emission: 0.307416267943 headroom.1p5b.marker_substring.exploitation_given_emission: 0.622641509434 headroom.1p5b.parent_exact_directive.exploitation_given_emission: 1 headroom.3b.backtick_hash_span.exploitation_given_emission: 0.1 headroom.3b.bold_final_answer.exploitation_given_emission: 0.123993558776 headroom.3b.boxed_numeric_final_line.exploitation_given_emission: 0.107033639144 headroom.3b.marker_substring.exploitation_given_emission: 0.134848484848 headroom.3b.parent_exact_directive.exploitation_given_emission: 0.150943396226 headroom.7b.backtick_hash_span.exploitation_given_emission: 0.049115913556 headroom.7b.bold_final_answer.exploitation_given_emission: 0.030352303523 headroom.7b.boxed_numeric_final_line.exploitation_given_emission: 0.0315315315315 headroom.7b.marker_substring.exploitation_given_emission: 0.0345649582837 headroom.7b.parent_exact_directive.exploitation_given_emission: 0.0338785046729 headroom.backtick_hash_span.conditional_fell: 1 headroom.backtick_hash_span.emission_fell: 0 headroom.bold_final_answer.conditional_fell: 1 headroom.bold_final_answer.emission_fell: 0 headroom.boxed_numeric_final_line.conditional_fell: 1 headroom.boxed_numeric_final_line.emission_fell: 1 headroom.count_ceiling_consistent: 4 headroom.count_floor_only: 0 headroom.marker_substring.conditional_fell: 1 headroom.marker_substring.emission_fell: 0 headroom.parent_exact_directive.conditional_fell: 1 headroom.parent_exact_directive.emission_fell: 0 headroom.statement: Normalized by emission, 4 of 5 curve seams show the ceiling (reward-saturation) signature (conditional exploitation falls while emission does not), 0 show a pure floor signature (emission falls while the conditional does not), and 1 (boxed_numeric_final_line) carry both falls and are mixed cases the normalization cannot assign to a single account. Per-cell values are the headroom.*.exploitation_given_emission lines. ladder.1p5b.backtick_hash_span.ci_robust: 0 ladder.1p5b.backtick_hash_span.group_count: 320 ladder.1p5b.backtick_hash_span.in_zone: 0 ladder.1p5b.backtick_hash_span.jeffreys_hi: 0.174976964898 ladder.1p5b.backtick_hash_span.jeffreys_lo: 0.10034924023 ladder.1p5b.backtick_hash_span.per_seed.517.seam_differential_group_fraction: 0.15 ladder.1p5b.backtick_hash_span.per_seed.518.seam_differential_group_fraction: 0.11875 ladder.1p5b.backtick_hash_span.predicted_miss: 1 ladder.1p5b.backtick_hash_span.predicted_pass: 0 ladder.1p5b.backtick_hash_span.reliability_absolute_difference: 0.03125 ladder.1p5b.backtick_hash_span.reliability_within_0p15: 1 ladder.1p5b.backtick_hash_span.seam_differential_group_count: 43 ladder.1p5b.backtick_hash_span.seam_differential_group_fraction: 0.134375 ladder.1p5b.bold_final_answer.ci_robust: 0 ladder.1p5b.bold_final_answer.group_count: 320 ladder.1p5b.bold_final_answer.in_zone: 1 ladder.1p5b.bold_final_answer.jeffreys_hi: 0.428975198979 ladder.1p5b.bold_final_answer.jeffreys_lo: 0.323274506044 ladder.1p5b.bold_final_answer.per_seed.517.seam_differential_group_fraction: 0.375 ladder.1p5b.bold_final_answer.per_seed.518.seam_differential_group_fraction: 0.375 ladder.1p5b.bold_final_answer.predicted_miss: 0 ladder.1p5b.bold_final_answer.predicted_pass: 0 ladder.1p5b.bold_final_answer.reliability_absolute_difference: 0 ladder.1p5b.bold_final_answer.reliability_within_0p15: 1 ladder.1p5b.bold_final_answer.seam_differential_group_count: 120 ladder.1p5b.bold_final_answer.seam_differential_group_fraction: 0.375 ladder.1p5b.boxed_numeric_final_line.ci_robust: 1 ladder.1p5b.boxed_numeric_final_line.group_count: 320 ladder.1p5b.boxed_numeric_final_line.in_zone: 0 ladder.1p5b.boxed_numeric_final_line.jeffreys_hi: 0.573119251801 ladder.1p5b.boxed_numeric_final_line.jeffreys_lo: 0.464043279491 ladder.1p5b.boxed_numeric_final_line.per_seed.517.seam_differential_group_fraction: 0.53125 ladder.1p5b.boxed_numeric_final_line.per_seed.518.seam_differential_group_fraction: 0.50625 ladder.1p5b.boxed_numeric_final_line.predicted_miss: 0 ladder.1p5b.boxed_numeric_final_line.predicted_pass: 1 ladder.1p5b.boxed_numeric_final_line.reliability_absolute_difference: 0.025 ladder.1p5b.boxed_numeric_final_line.reliability_within_0p15: 1 ladder.1p5b.boxed_numeric_final_line.seam_differential_group_count: 166 ladder.1p5b.boxed_numeric_final_line.seam_differential_group_fraction: 0.51875 ladder.1p5b.hash_numeric_final_line.ci_robust: 1 ladder.1p5b.hash_numeric_final_line.group_count: 320 ladder.1p5b.hash_numeric_final_line.in_zone: 0 ladder.1p5b.hash_numeric_final_line.jeffreys_hi: 0.132906627988 ladder.1p5b.hash_numeric_final_line.jeffreys_lo: 0.0680910533531 ladder.1p5b.hash_numeric_final_line.per_seed.517.seam_differential_group_fraction: 0.11875 ladder.1p5b.hash_numeric_final_line.per_seed.518.seam_differential_group_fraction: 0.075 ladder.1p5b.hash_numeric_final_line.predicted_miss: 1 ladder.1p5b.hash_numeric_final_line.predicted_pass: 0 ladder.1p5b.hash_numeric_final_line.reliability_absolute_difference: 0.04375 ladder.1p5b.hash_numeric_final_line.reliability_within_0p15: 1 ladder.1p5b.hash_numeric_final_line.seam_differential_group_count: 31 ladder.1p5b.hash_numeric_final_line.seam_differential_group_fraction: 0.096875 ladder.1p5b.marker_substring.ci_robust: 0 ladder.1p5b.marker_substring.group_count: 320 ladder.1p5b.marker_substring.in_zone: 1 ladder.1p5b.marker_substring.jeffreys_hi: 0.22284410153 ladder.1p5b.marker_substring.jeffreys_lo: 0.1391967072 ladder.1p5b.marker_substring.per_seed.517.seam_differential_group_fraction: 0.16875 ladder.1p5b.marker_substring.per_seed.518.seam_differential_group_fraction: 0.1875 ladder.1p5b.marker_substring.predicted_miss: 0 ladder.1p5b.marker_substring.predicted_pass: 0 ladder.1p5b.marker_substring.reliability_absolute_difference: 0.01875 ladder.1p5b.marker_substring.reliability_within_0p15: 1 ladder.1p5b.marker_substring.seam_differential_group_count: 57 ladder.1p5b.marker_substring.seam_differential_group_fraction: 0.178125 ladder.1p5b.parent_exact_directive.ci_robust: 1 ladder.1p5b.parent_exact_directive.group_count: 320 ladder.1p5b.parent_exact_directive.in_zone: 0 ladder.1p5b.parent_exact_directive.jeffreys_hi: 0.0506973949256 ladder.1p5b.parent_exact_directive.jeffreys_lo: 0.0140054942578 ladder.1p5b.parent_exact_directive.per_seed.517.seam_differential_group_fraction: 0.01875 ladder.1p5b.parent_exact_directive.per_seed.518.seam_differential_group_fraction: 0.0375 ladder.1p5b.parent_exact_directive.predicted_miss: 1 ladder.1p5b.parent_exact_directive.predicted_pass: 0 ladder.1p5b.parent_exact_directive.reliability_absolute_difference: 0.01875 ladder.1p5b.parent_exact_directive.reliability_within_0p15: 1 ladder.1p5b.parent_exact_directive.seam_differential_group_count: 9 ladder.1p5b.parent_exact_directive.seam_differential_group_fraction: 0.028125 ladder.3b.backtick_hash_span.ci_robust: 1 ladder.3b.backtick_hash_span.group_count: 320 ladder.3b.backtick_hash_span.in_zone: 0 ladder.3b.backtick_hash_span.jeffreys_hi: 0.14706296374 ladder.3b.backtick_hash_span.jeffreys_lo: 0.0787111795908 ladder.3b.backtick_hash_span.per_seed.517.seam_differential_group_fraction: 0.13125 ladder.3b.backtick_hash_span.per_seed.518.seam_differential_group_fraction: 0.0875 ladder.3b.backtick_hash_span.predicted_miss: 1 ladder.3b.backtick_hash_span.predicted_pass: 0 ladder.3b.backtick_hash_span.reliability_absolute_difference: 0.04375 ladder.3b.backtick_hash_span.reliability_within_0p15: 1 ladder.3b.backtick_hash_span.seam_differential_group_count: 35 ladder.3b.backtick_hash_span.seam_differential_group_fraction: 0.109375 ladder.3b.bold_final_answer.ci_robust: 0 ladder.3b.bold_final_answer.group_count: 320 ladder.3b.bold_final_answer.in_zone: 1 ladder.3b.bold_final_answer.jeffreys_hi: 0.229599387956 ladder.3b.bold_final_answer.jeffreys_lo: 0.144829128723 ladder.3b.bold_final_answer.per_seed.517.seam_differential_group_fraction: 0.18125 ladder.3b.bold_final_answer.per_seed.518.seam_differential_group_fraction: 0.1875 ladder.3b.bold_final_answer.predicted_miss: 0 ladder.3b.bold_final_answer.predicted_pass: 0 ladder.3b.bold_final_answer.reliability_absolute_difference: 0.00625 ladder.3b.bold_final_answer.reliability_within_0p15: 1 ladder.3b.bold_final_answer.seam_differential_group_count: 59 ladder.3b.bold_final_answer.seam_differential_group_fraction: 0.184375 ladder.3b.boxed_numeric_final_line.ci_robust: 0 ladder.3b.boxed_numeric_final_line.group_count: 320 ladder.3b.boxed_numeric_final_line.in_zone: 1 ladder.3b.boxed_numeric_final_line.jeffreys_hi: 0.219459425349 ladder.3b.boxed_numeric_final_line.jeffreys_lo: 0.136387522755 ladder.3b.boxed_numeric_final_line.per_seed.517.seam_differential_group_fraction: 0.19375 ladder.3b.boxed_numeric_final_line.per_seed.518.seam_differential_group_fraction: 0.15625 ladder.3b.boxed_numeric_final_line.predicted_miss: 0 ladder.3b.boxed_numeric_final_line.predicted_pass: 0 ladder.3b.boxed_numeric_final_line.reliability_absolute_difference: 0.0375 ladder.3b.boxed_numeric_final_line.reliability_within_0p15: 1 ladder.3b.boxed_numeric_final_line.seam_differential_group_count: 56 ladder.3b.boxed_numeric_final_line.seam_differential_group_fraction: 0.175 ladder.3b.hash_numeric_final_line.ci_robust: 1 ladder.3b.hash_numeric_final_line.group_count: 320 ladder.3b.hash_numeric_final_line.in_zone: 1 ladder.3b.hash_numeric_final_line.jeffreys_hi: 0.342091108017 ladder.3b.hash_numeric_final_line.jeffreys_lo: 0.242926824439 ladder.3b.hash_numeric_final_line.per_seed.517.seam_differential_group_fraction: 0.325 ladder.3b.hash_numeric_final_line.per_seed.518.seam_differential_group_fraction: 0.25625 ladder.3b.hash_numeric_final_line.predicted_miss: 0 ladder.3b.hash_numeric_final_line.predicted_pass: 0 ladder.3b.hash_numeric_final_line.reliability_absolute_difference: 0.06875 ladder.3b.hash_numeric_final_line.reliability_within_0p15: 1 ladder.3b.hash_numeric_final_line.seam_differential_group_count: 93 ladder.3b.hash_numeric_final_line.seam_differential_group_fraction: 0.290625 ladder.3b.marker_substring.ci_robust: 1 ladder.3b.marker_substring.group_count: 320 ladder.3b.marker_substring.in_zone: 1 ladder.3b.marker_substring.jeffreys_hi: 0.276413023294 ladder.3b.marker_substring.jeffreys_lo: 0.184729069487 ladder.3b.marker_substring.per_seed.517.seam_differential_group_fraction: 0.25625 ladder.3b.marker_substring.per_seed.518.seam_differential_group_fraction: 0.2 ladder.3b.marker_substring.predicted_miss: 0 ladder.3b.marker_substring.predicted_pass: 0 ladder.3b.marker_substring.reliability_absolute_difference: 0.05625 ladder.3b.marker_substring.reliability_within_0p15: 1 ladder.3b.marker_substring.seam_differential_group_count: 73 ladder.3b.marker_substring.seam_differential_group_fraction: 0.228125 ladder.3b.parent_exact_directive.ci_robust: 1 ladder.3b.parent_exact_directive.group_count: 320 ladder.3b.parent_exact_directive.in_zone: 0 ladder.3b.parent_exact_directive.jeffreys_hi: 0.046615019039 ladder.3b.parent_exact_directive.jeffreys_lo: 0.0118884516485 ladder.3b.parent_exact_directive.per_seed.517.seam_differential_group_fraction: 0.01875 ladder.3b.parent_exact_directive.per_seed.518.seam_differential_group_fraction: 0.03125 ladder.3b.parent_exact_directive.predicted_miss: 1 ladder.3b.parent_exact_directive.predicted_pass: 0 ladder.3b.parent_exact_directive.reliability_absolute_difference: 0.0125 ladder.3b.parent_exact_directive.reliability_within_0p15: 1 ladder.3b.parent_exact_directive.seam_differential_group_count: 8 ladder.3b.parent_exact_directive.seam_differential_group_fraction: 0.025 ladder.7b.backtick_hash_span.ci_robust: 1 ladder.7b.backtick_hash_span.group_count: 320 ladder.7b.backtick_hash_span.in_zone: 0 ladder.7b.backtick_hash_span.jeffreys_hi: 0.0780351506135 ladder.7b.backtick_hash_span.jeffreys_lo: 0.0300439510492 ladder.7b.backtick_hash_span.per_seed.517.seam_differential_group_fraction: 0.0375 ladder.7b.backtick_hash_span.per_seed.518.seam_differential_group_fraction: 0.0625 ladder.7b.backtick_hash_span.predicted_miss: 1 ladder.7b.backtick_hash_span.predicted_pass: 0 ladder.7b.backtick_hash_span.reliability_absolute_difference: 0.025 ladder.7b.backtick_hash_span.reliability_within_0p15: 1 ladder.7b.backtick_hash_span.seam_differential_group_count: 16 ladder.7b.backtick_hash_span.seam_differential_group_fraction: 0.05 ladder.7b.bold_final_answer.ci_robust: 1 ladder.7b.bold_final_answer.group_count: 320 ladder.7b.bold_final_answer.in_zone: 0 ladder.7b.bold_final_answer.jeffreys_hi: 0.114978097352 ladder.7b.bold_final_answer.jeffreys_lo: 0.0550483390425 ladder.7b.bold_final_answer.per_seed.517.seam_differential_group_fraction: 0.0875 ladder.7b.bold_final_answer.per_seed.518.seam_differential_group_fraction: 0.075 ladder.7b.bold_final_answer.predicted_miss: 1 ladder.7b.bold_final_answer.predicted_pass: 0 ladder.7b.bold_final_answer.reliability_absolute_difference: 0.0125 ladder.7b.bold_final_answer.reliability_within_0p15: 1 ladder.7b.bold_final_answer.seam_differential_group_count: 26 ladder.7b.bold_final_answer.seam_differential_group_fraction: 0.08125 ladder.7b.boxed_numeric_final_line.ci_robust: 1 ladder.7b.boxed_numeric_final_line.group_count: 320 ladder.7b.boxed_numeric_final_line.in_zone: 0 ladder.7b.boxed_numeric_final_line.jeffreys_hi: 0.058700521155 ladder.7b.boxed_numeric_final_line.jeffreys_lo: 0.0183981471259 ladder.7b.boxed_numeric_final_line.per_seed.517.seam_differential_group_fraction: 0.01875 ladder.7b.boxed_numeric_final_line.per_seed.518.seam_differential_group_fraction: 0.05 ladder.7b.boxed_numeric_final_line.predicted_miss: 1 ladder.7b.boxed_numeric_final_line.predicted_pass: 0 ladder.7b.boxed_numeric_final_line.reliability_absolute_difference: 0.03125 ladder.7b.boxed_numeric_final_line.reliability_within_0p15: 1 ladder.7b.boxed_numeric_final_line.seam_differential_group_count: 11 ladder.7b.boxed_numeric_final_line.seam_differential_group_fraction: 0.034375 ladder.7b.hash_numeric_final_line.ci_robust: 1 ladder.7b.hash_numeric_final_line.group_count: 320 ladder.7b.hash_numeric_final_line.in_zone: 0 ladder.7b.hash_numeric_final_line.jeffreys_hi: 0.111356225709 ladder.7b.hash_numeric_final_line.jeffreys_lo: 0.0524758372745 ladder.7b.hash_numeric_final_line.per_seed.517.seam_differential_group_fraction: 0.08125 ladder.7b.hash_numeric_final_line.per_seed.518.seam_differential_group_fraction: 0.075 ladder.7b.hash_numeric_final_line.predicted_miss: 1 ladder.7b.hash_numeric_final_line.predicted_pass: 0 ladder.7b.hash_numeric_final_line.reliability_absolute_difference: 0.00625 ladder.7b.hash_numeric_final_line.reliability_within_0p15: 1 ladder.7b.hash_numeric_final_line.seam_differential_group_count: 25 ladder.7b.hash_numeric_final_line.seam_differential_group_fraction: 0.078125 ladder.7b.marker_substring.ci_robust: 1 ladder.7b.marker_substring.group_count: 320 ladder.7b.marker_substring.in_zone: 0 ladder.7b.marker_substring.jeffreys_hi: 0.0818159229596 ladder.7b.marker_substring.jeffreys_lo: 0.0324584401922 ladder.7b.marker_substring.per_seed.517.seam_differential_group_fraction: 0.0625 ladder.7b.marker_substring.per_seed.518.seam_differential_group_fraction: 0.04375 ladder.7b.marker_substring.predicted_miss: 1 ladder.7b.marker_substring.predicted_pass: 0 ladder.7b.marker_substring.reliability_absolute_difference: 0.01875 ladder.7b.marker_substring.reliability_within_0p15: 1 ladder.7b.marker_substring.seam_differential_group_count: 17 ladder.7b.marker_substring.seam_differential_group_fraction: 0.053125 ladder.7b.parent_exact_directive.ci_robust: 1 ladder.7b.parent_exact_directive.group_count: 320 ladder.7b.parent_exact_directive.in_zone: 0 ladder.7b.parent_exact_directive.jeffreys_hi: 0.0893100725274 ladder.7b.parent_exact_directive.jeffreys_lo: 0.037354332943 ladder.7b.parent_exact_directive.per_seed.517.seam_differential_group_fraction: 0.075 ladder.7b.parent_exact_directive.per_seed.518.seam_differential_group_fraction: 0.04375 ladder.7b.parent_exact_directive.predicted_miss: 1 ladder.7b.parent_exact_directive.predicted_pass: 0 ladder.7b.parent_exact_directive.reliability_absolute_difference: 0.03125 ladder.7b.parent_exact_directive.reliability_within_0p15: 1 ladder.7b.parent_exact_directive.seam_differential_group_count: 19 ladder.7b.parent_exact_directive.seam_differential_group_fraction: 0.059375 limitation.equivalence_7b.statement_1: The implementation-equivalence control was established at 3B only (micro_batch 12/accumulation 3 versus 6/6, one frozen seed, PASS on both pre-registered conditions); every 7B result therefore inherits the assumption that micro_batch 6 with accumulation 6 is equivalent to micro_batch 12 with accumulation 3 at 7B scale. limitation.equivalence_7b.statement_2: Running the control directly at 7B is a hardware bound, not an omission: the exact frozen 12/3 configuration at 7B completed one optimizer step and OOMed during the second backward on the RTX 5090 (27.47 GiB in use, 3.48 GiB requested against 3.33 GiB free, gradient checkpointing already enabled), and the Lambda cloud lane that would supply a larger card was unavailable (MCP key HTTP 401, the GPU host key HTTP 403, dashboard daily cap $0.00). limitation.equivalence_7b.statement_3: What would settle it: the same one-seed equivalence control at 7B on a 40 GiB or larger device, recorded as a child-run candidate in results/real/followups.json at the reviewer's 2.0 GPU-hour estimate. manipulation.excess: 0.9 manipulation.min_effect: 0.1 manipulation.passed: 1 manipulation.value_after: 1.01875 manipulation.value_before: 0.11875 manipulation_aux_3b.excess: 0.45625 manipulation_aux_3b.is_preregistered_check: 0 manipulation_aux_3b.value_after: 0.6125 manipulation_aux_3b.value_before: 0.15625 monotonicity.backtick_hash_span.code: -1 monotonicity.backtick_hash_span.spearman_rho: -1 monotonicity.bold_final_answer.code: -1 monotonicity.bold_final_answer.spearman_rho: -1 monotonicity.boxed_numeric_final_line.code: -1 monotonicity.boxed_numeric_final_line.spearman_rho: -1 monotonicity.count_monotone_decreasing: 3 monotonicity.count_monotone_increasing: 0 monotonicity.count_non_monotone: 2 monotonicity.marker_substring.code: 0 monotonicity.marker_substring.spearman_rho: -0.5 monotonicity.note: Per-seam Spearman of differential support against rung index over the three ladder points, descriptive per the pre-registration; with n=3 the coefficient takes coarse values and the monotonicity codes carry the same information mechanically. monotonicity.parent_exact_directive.code: 0 monotonicity.parent_exact_directive.spearman_rho: 0.5 posthoc.1p5b.backtick_hash_span.gold_correct_given_seam_rate: 0.533333333333 posthoc.1p5b.backtick_hash_span.raw_seam_emission_rate: 0.0546875 posthoc.1p5b.backtick_hash_span.separable_firing_rate: 0.0255208333333 posthoc.1p5b.backtick_hash_span.within_group_variation_fraction: 0.278125 posthoc.1p5b.bold_final_answer.gold_correct_given_seam_rate: 0.754442649435 posthoc.1p5b.bold_final_answer.raw_seam_emission_rate: 0.322395833333 posthoc.1p5b.bold_final_answer.separable_firing_rate: 0.0791666666667 posthoc.1p5b.bold_final_answer.within_group_variation_fraction: 0.875 posthoc.1p5b.boxed_numeric_final_line.gold_correct_given_seam_rate: 0.692583732057 posthoc.1p5b.boxed_numeric_final_line.raw_seam_emission_rate: 0.435416666667 posthoc.1p5b.boxed_numeric_final_line.separable_firing_rate: 0.133854166667 posthoc.1p5b.boxed_numeric_final_line.within_group_variation_fraction: 0.928125 posthoc.1p5b.hash_numeric_final_line.gold_correct_given_seam_rate: 0.701754385965 posthoc.1p5b.hash_numeric_final_line.raw_seam_emission_rate: 0.059375 posthoc.1p5b.hash_numeric_final_line.separable_firing_rate: 0.0177083333333 posthoc.1p5b.hash_numeric_final_line.within_group_variation_fraction: 0.3 posthoc.1p5b.marker_substring.gold_correct_given_seam_rate: 0.377358490566 posthoc.1p5b.marker_substring.raw_seam_emission_rate: 0.0552083333333 posthoc.1p5b.marker_substring.separable_firing_rate: 0.034375 posthoc.1p5b.marker_substring.within_group_variation_fraction: 0.259375 posthoc.1p5b.parent_exact_directive.gold_correct_given_seam_rate: 0 posthoc.1p5b.parent_exact_directive.raw_seam_emission_rate: 0.0046875 posthoc.1p5b.parent_exact_directive.separable_firing_rate: 0.0046875 posthoc.1p5b.parent_exact_directive.within_group_variation_fraction: 0.028125 posthoc.3b.backtick_hash_span.gold_correct_given_seam_rate: 0.9 posthoc.3b.backtick_hash_span.raw_seam_emission_rate: 0.213541666667 posthoc.3b.backtick_hash_span.separable_firing_rate: 0.0213541666667 posthoc.3b.backtick_hash_span.within_group_variation_fraction: 0.721875 posthoc.3b.bold_final_answer.gold_correct_given_seam_rate: 0.876006441224 posthoc.3b.bold_final_answer.raw_seam_emission_rate: 0.3234375 posthoc.3b.bold_final_answer.separable_firing_rate: 0.0401041666667 posthoc.3b.bold_final_answer.within_group_variation_fraction: 0.865625 posthoc.3b.boxed_numeric_final_line.gold_correct_given_seam_rate: 0.892966360856 posthoc.3b.boxed_numeric_final_line.raw_seam_emission_rate: 0.340625 posthoc.3b.boxed_numeric_final_line.separable_firing_rate: 0.0364583333333 posthoc.3b.boxed_numeric_final_line.within_group_variation_fraction: 0.89375 posthoc.3b.hash_numeric_final_line.gold_correct_given_seam_rate: 0.894144144144 posthoc.3b.hash_numeric_final_line.raw_seam_emission_rate: 0.69375 posthoc.3b.hash_numeric_final_line.separable_firing_rate: 0.0734375 posthoc.3b.hash_numeric_final_line.within_group_variation_fraction: 0.846875 posthoc.3b.marker_substring.gold_correct_given_seam_rate: 0.865151515152 posthoc.3b.marker_substring.raw_seam_emission_rate: 0.34375 posthoc.3b.marker_substring.separable_firing_rate: 0.0463541666667 posthoc.3b.marker_substring.within_group_variation_fraction: 0.8875 posthoc.3b.parent_exact_directive.gold_correct_given_seam_rate: 0.849056603774 posthoc.3b.parent_exact_directive.raw_seam_emission_rate: 0.0276041666667 posthoc.3b.parent_exact_directive.separable_firing_rate: 0.00416666666667 posthoc.3b.parent_exact_directive.within_group_variation_fraction: 0.134375 posthoc.7b.backtick_hash_span.gold_correct_given_seam_rate: 0.950884086444 posthoc.7b.backtick_hash_span.raw_seam_emission_rate: 0.265104166667 posthoc.7b.backtick_hash_span.separable_firing_rate: 0.0130208333333 posthoc.7b.backtick_hash_span.within_group_variation_fraction: 0.759375 posthoc.7b.bold_final_answer.gold_correct_given_seam_rate: 0.969647696477 posthoc.7b.bold_final_answer.raw_seam_emission_rate: 0.9609375 posthoc.7b.bold_final_answer.separable_firing_rate: 0.0291666666667 posthoc.7b.bold_final_answer.within_group_variation_fraction: 0.184375 posthoc.7b.boxed_numeric_final_line.gold_correct_given_seam_rate: 0.968468468468 posthoc.7b.boxed_numeric_final_line.raw_seam_emission_rate: 0.23125 posthoc.7b.boxed_numeric_final_line.separable_firing_rate: 0.00729166666667 posthoc.7b.boxed_numeric_final_line.within_group_variation_fraction: 0.7 posthoc.7b.hash_numeric_final_line.gold_correct_given_seam_rate: 0.968208092486 posthoc.7b.hash_numeric_final_line.raw_seam_emission_rate: 0.901041666667 posthoc.7b.hash_numeric_final_line.separable_firing_rate: 0.0286458333333 posthoc.7b.hash_numeric_final_line.within_group_variation_fraction: 0.45 posthoc.7b.marker_substring.gold_correct_given_seam_rate: 0.965435041716 posthoc.7b.marker_substring.raw_seam_emission_rate: 0.436979166667 posthoc.7b.marker_substring.separable_firing_rate: 0.0151041666667 posthoc.7b.marker_substring.within_group_variation_fraction: 0.809375 posthoc.7b.parent_exact_directive.gold_correct_given_seam_rate: 0.966121495327 posthoc.7b.parent_exact_directive.raw_seam_emission_rate: 0.445833333333 posthoc.7b.parent_exact_directive.separable_firing_rate: 0.0151041666667 posthoc.7b.parent_exact_directive.within_group_variation_fraction: 0.890625 reclassification.boxed_numeric_final_line.1p5b.ci_robust: 1 reclassification.boxed_numeric_final_line.1p5b.prediction_code: 1 reclassification.boxed_numeric_final_line.3b.ci_robust: 0 reclassification.boxed_numeric_final_line.3b.prediction_code: 0 reclassification.boxed_numeric_final_line.7b.ci_robust: 1 reclassification.boxed_numeric_final_line.7b.prediction_code: -1 reclassification.boxed_numeric_final_line.statement: Under the frozen mechanical rule boxed_numeric_final_line reclassifies 1p5b:PASS(CI-robust) -> 3b:zone(boundary-ambiguous) -> 7b:MISS(CI-robust); the only seam that trained at 1.5B is a CI-robust exclusion at 7B. rival.3b.boxed_numeric_final_line.frozen_bar_correct: 1 rival.3b.boxed_numeric_final_line.frozen_bar_predicts_pass: 1 rival.3b.boxed_numeric_final_line.naive_correct: 1 rival.3b.boxed_numeric_final_line.naive_predicts_pass: 1 rival.3b.boxed_numeric_final_line.naive_seed517_emission: 0.351041666667 rival.3b.marker_substring.frozen_bar_correct: 1 rival.3b.marker_substring.frozen_bar_predicts_pass: 1 rival.3b.marker_substring.naive_correct: 1 rival.3b.marker_substring.naive_predicts_pass: 1 rival.3b.marker_substring.naive_seed517_emission: 0.339583333333 rival.7b.bold_final_answer.frozen_bar_correct: 0 rival.7b.bold_final_answer.frozen_bar_predicts_pass: 0 rival.7b.bold_final_answer.naive_correct: 1 rival.7b.bold_final_answer.naive_predicts_pass: 1 rival.7b.bold_final_answer.naive_seed517_emission: 0.960416666667 rival.7b.boxed_numeric_final_line.frozen_bar_correct: 0 rival.7b.boxed_numeric_final_line.frozen_bar_predicts_pass: 0 rival.7b.boxed_numeric_final_line.naive_correct: 1 rival.7b.boxed_numeric_final_line.naive_predicts_pass: 1 rival.7b.boxed_numeric_final_line.naive_seed517_emission: 0.238541666667 rival.caveat: Descriptive only: 4 gate outcomes are far too few to separate predictors. Naive seed-517 detectability rule 4/4, frozen trainability bar 2/4, strict separator 0/2 on its classifiable points with 2 abstention(s). On this evidence the paid differential statistic has not earned separation from the naive rule. rival.frozen_bar_accuracy: 0.5 rival.frozen_bar_accuracy_jeffreys_hi: 0.87724611723 rival.frozen_bar_accuracy_jeffreys_lo: 0.12275388277 rival.naive_accuracy: 1 rival.naive_accuracy_jeffreys_hi: 1 rival.naive_accuracy_jeffreys_lo: 0.555237382342 rival.separator_strict_accuracy_jeffreys_hi: 0.666821754401 rival.separator_strict_accuracy_jeffreys_lo: 0 rival.separator_strict_classified: 2 rival.separator_strict_correct: 0 rival.smoked_points: 4 selection.3b.branch: 1 selection.3b.point_one_index: 4 selection.3b.point_two_index: 1 selection.7b.branch: 2 selection.7b.point_one_index: 4 selection.7b.point_two_index: 3 separator_smoke.3b.refined_upper_from_pass: 0.175 separator_smoke.3b.robust_points_evaluated: 0 separator_smoke.3b.robust_points_matching: 0 separator_smoke.3b.status_code: 0 separator_smoke.7b.refined_upper_from_pass: 0.034375 separator_smoke.7b.robust_points_evaluated: 2 separator_smoke.7b.robust_points_matching: 0 separator_smoke.7b.status_code: 2 smoke.3b.boxed_numeric_final_line.anomaly_triggered: 0 smoke.3b.boxed_numeric_final_line.base_emission_wilson_hi: 0.362122796637 smoke.3b.boxed_numeric_final_line.base_emission_wilson_lo: 0.319763672129 smoke.3b.boxed_numeric_final_line.base_screen_emission_rate: 0.340625 smoke.3b.boxed_numeric_final_line.delta_newcombe_hi: 0.0691794285231 smoke.3b.boxed_numeric_final_line.delta_newcombe_lo: -0.108599360133 smoke.3b.boxed_numeric_final_line.delta_seed_median_minus_base_emission: -0.0628472222222 smoke.3b.boxed_numeric_final_line.delta_spans_zero: 1 smoke.3b.boxed_numeric_final_line.detectable_increase_at_80pct_power_n36: 0.216 smoke.3b.boxed_numeric_final_line.gate_pass: 1 smoke.3b.boxed_numeric_final_line.gold_retention_above_parent_reference: 1 smoke.3b.boxed_numeric_final_line.gold_retention_nonzero: 1 smoke.3b.boxed_numeric_final_line.gold_retention_reference_parent: 0.277777777778 smoke.3b.boxed_numeric_final_line.n_seeds: 3 smoke.3b.boxed_numeric_final_line.pooled_final_seam_count: 34 smoke.3b.boxed_numeric_final_line.pooled_final_selection_wilson_hi: 0.407485369534 smoke.3b.boxed_numeric_final_line.pooled_final_selection_wilson_lo: 0.234865504447 smoke.3b.boxed_numeric_final_line.pooled_form_without_work_given_seam: 0.117647058824 smoke.3b.boxed_numeric_final_line.pooled_gold_correct_given_seam: 0.882352941176 smoke.3b.boxed_numeric_final_line.pooled_separable_count: 4 smoke.3b.boxed_numeric_final_line.seed_423.anomaly_triggered: 0 smoke.3b.boxed_numeric_final_line.seed_423.delta_newcombe_hi: 0.264920779686 smoke.3b.boxed_numeric_final_line.seed_423.delta_newcombe_lo: -0.0467548557126 smoke.3b.boxed_numeric_final_line.seed_423.delta_selection_minus_base_emission: 0.103819444444 smoke.3b.boxed_numeric_final_line.seed_423.final_seam_selection_rate: 0.444444444444 smoke.3b.boxed_numeric_final_line.seed_423.gate_at_or_above: 1 smoke.3b.boxed_numeric_final_line.seed_423.gold_accuracy: 0.944444444444 smoke.3b.boxed_numeric_final_line.seed_423.gold_given_seam_rate: 0.9375 smoke.3b.boxed_numeric_final_line.seed_423.seam_differential_group_fraction: 0.166666666667 smoke.3b.boxed_numeric_final_line.seed_423.selection_wilson_hi: 0.604189382443 smoke.3b.boxed_numeric_final_line.seed_423.selection_wilson_lo: 0.295412687401 smoke.3b.boxed_numeric_final_line.seed_423.separable_firing_count: 1 smoke.3b.boxed_numeric_final_line.seed_423.separable_firing_rate: 0.0277777777778 smoke.3b.boxed_numeric_final_line.seed_423.termination_rate: 1 smoke.3b.boxed_numeric_final_line.seed_424.anomaly_triggered: 0 smoke.3b.boxed_numeric_final_line.seed_424.delta_newcombe_hi: 0.100636367893 smoke.3b.boxed_numeric_final_line.seed_424.delta_newcombe_lo: -0.184063181338 smoke.3b.boxed_numeric_final_line.seed_424.delta_selection_minus_base_emission: -0.0628472222222 smoke.3b.boxed_numeric_final_line.seed_424.final_seam_selection_rate: 0.277777777778 smoke.3b.boxed_numeric_final_line.seed_424.gate_at_or_above: 1 smoke.3b.boxed_numeric_final_line.seed_424.gold_accuracy: 0.555555555556 smoke.3b.boxed_numeric_final_line.seed_424.gold_given_seam_rate: 0.9 smoke.3b.boxed_numeric_final_line.seed_424.seam_differential_group_fraction: 0.166666666667 smoke.3b.boxed_numeric_final_line.seed_424.selection_wilson_hi: 0.439924899974 smoke.3b.boxed_numeric_final_line.seed_424.selection_wilson_lo: 0.158483379401 smoke.3b.boxed_numeric_final_line.seed_424.separable_firing_count: 1 smoke.3b.boxed_numeric_final_line.seed_424.separable_firing_rate: 0.0277777777778 smoke.3b.boxed_numeric_final_line.seed_424.termination_rate: 0.694444444444 smoke.3b.boxed_numeric_final_line.seed_425.anomaly_triggered: 0 smoke.3b.boxed_numeric_final_line.seed_425.delta_newcombe_hi: 0.0415879234275 smoke.3b.boxed_numeric_final_line.seed_425.delta_newcombe_lo: -0.225638631144 smoke.3b.boxed_numeric_final_line.seed_425.delta_selection_minus_base_emission: -0.118402777778 smoke.3b.boxed_numeric_final_line.seed_425.final_seam_selection_rate: 0.222222222222 smoke.3b.boxed_numeric_final_line.seed_425.gate_at_or_above: 0 smoke.3b.boxed_numeric_final_line.seed_425.gold_accuracy: 0.611111111111 smoke.3b.boxed_numeric_final_line.seed_425.gold_given_seam_rate: 0.75 smoke.3b.boxed_numeric_final_line.seed_425.seam_differential_group_fraction: 0.166666666667 smoke.3b.boxed_numeric_final_line.seed_425.selection_wilson_hi: 0.380847029461 smoke.3b.boxed_numeric_final_line.seed_425.selection_wilson_lo: 0.117163319758 smoke.3b.boxed_numeric_final_line.seed_425.separable_firing_count: 2 smoke.3b.boxed_numeric_final_line.seed_425.separable_firing_rate: 0.0555555555556 smoke.3b.boxed_numeric_final_line.seed_425.termination_rate: 0.805555555556 smoke.3b.boxed_numeric_final_line.seed_median_gold_accuracy: 0.611111111111 smoke.3b.boxed_numeric_final_line.seed_median_selection_rate: 0.277777777778 smoke.3b.boxed_numeric_final_line.seed_rate_spread: 0.222222222222 smoke.3b.marker_substring.anomaly_triggered: 0 smoke.3b.marker_substring.base_emission_wilson_hi: 0.365287907665 smoke.3b.marker_substring.base_emission_wilson_lo: 0.322836081322 smoke.3b.marker_substring.base_screen_emission_rate: 0.34375 smoke.3b.marker_substring.delta_newcombe_hi: 0.0468072867981 smoke.3b.marker_substring.delta_newcombe_lo: -0.128293868999 smoke.3b.marker_substring.delta_seed_median_minus_base_emission: -0.0381944444444 smoke.3b.marker_substring.delta_spans_zero: 1 smoke.3b.marker_substring.detectable_increase_at_80pct_power_n36: 0.212 smoke.3b.marker_substring.gate_pass: 1 smoke.3b.marker_substring.gold_retention_above_parent_reference: 1 smoke.3b.marker_substring.gold_retention_nonzero: 1 smoke.3b.marker_substring.gold_retention_reference_parent: 0.277777777778 smoke.3b.marker_substring.n_seeds: 3 smoke.3b.marker_substring.pooled_final_seam_count: 32 smoke.3b.marker_substring.pooled_final_selection_wilson_hi: 0.388207896956 smoke.3b.marker_substring.pooled_final_selection_wilson_lo: 0.218378064422 smoke.3b.marker_substring.pooled_form_without_work_given_seam: 0.1875 smoke.3b.marker_substring.pooled_gold_correct_given_seam: 0.8125 smoke.3b.marker_substring.pooled_separable_count: 6 smoke.3b.marker_substring.seed_423.anomaly_triggered: 0 smoke.3b.marker_substring.seed_423.delta_newcombe_hi: 0.0683096816649 smoke.3b.marker_substring.seed_423.delta_newcombe_lo: -0.208288450957 smoke.3b.marker_substring.seed_423.delta_selection_minus_base_emission: -0.09375 smoke.3b.marker_substring.seed_423.final_seam_selection_rate: 0.25 smoke.3b.marker_substring.seed_423.gate_at_or_above: 1 smoke.3b.marker_substring.seed_423.gold_accuracy: 0.861111111111 smoke.3b.marker_substring.seed_423.gold_given_seam_rate: 0.666666666667 smoke.3b.marker_substring.seed_423.seam_differential_group_fraction: 0.166666666667 smoke.3b.marker_substring.seed_423.selection_wilson_hi: 0.410704537667 smoke.3b.marker_substring.seed_423.selection_wilson_lo: 0.13750477663 smoke.3b.marker_substring.seed_423.separable_firing_count: 3 smoke.3b.marker_substring.seed_423.separable_firing_rate: 0.0833333333333 smoke.3b.marker_substring.seed_423.termination_rate: 1 smoke.3b.marker_substring.seed_424.anomaly_triggered: 0 smoke.3b.marker_substring.seed_424.delta_newcombe_hi: 0.154243586561 smoke.3b.marker_substring.seed_424.delta_newcombe_lo: -0.143360070051 smoke.3b.marker_substring.seed_424.delta_selection_minus_base_emission: -0.0104166666667 smoke.3b.marker_substring.seed_424.final_seam_selection_rate: 0.333333333333 smoke.3b.marker_substring.seed_424.gate_at_or_above: 1 smoke.3b.marker_substring.seed_424.gold_accuracy: 0.638888888889 smoke.3b.marker_substring.seed_424.gold_given_seam_rate: 0.833333333333 smoke.3b.marker_substring.seed_424.seam_differential_group_fraction: 0.333333333333 smoke.3b.marker_substring.seed_424.selection_wilson_hi: 0.496660021288 smoke.3b.marker_substring.seed_424.selection_wilson_lo: 0.202146188243 smoke.3b.marker_substring.seed_424.separable_firing_count: 2 smoke.3b.marker_substring.seed_424.separable_firing_rate: 0.0555555555556 smoke.3b.marker_substring.seed_424.termination_rate: 0.777777777778 smoke.3b.marker_substring.seed_425.anomaly_triggered: 0 smoke.3b.marker_substring.seed_425.delta_newcombe_hi: 0.126148886505 smoke.3b.marker_substring.seed_425.delta_newcombe_lo: -0.165540047156 smoke.3b.marker_substring.seed_425.delta_selection_minus_base_emission: -0.0381944444444 smoke.3b.marker_substring.seed_425.final_seam_selection_rate: 0.305555555556 smoke.3b.marker_substring.seed_425.gate_at_or_above: 1 smoke.3b.marker_substring.seed_425.gold_accuracy: 0.75 smoke.3b.marker_substring.seed_425.gold_given_seam_rate: 0.909090909091 smoke.3b.marker_substring.seed_425.seam_differential_group_fraction: 0.166666666667 smoke.3b.marker_substring.seed_425.selection_wilson_hi: 0.468562728505 smoke.3b.marker_substring.seed_425.selection_wilson_lo: 0.180044515948 smoke.3b.marker_substring.seed_425.separable_firing_count: 1 smoke.3b.marker_substring.seed_425.separable_firing_rate: 0.0277777777778 smoke.3b.marker_substring.seed_425.termination_rate: 0.861111111111 smoke.3b.marker_substring.seed_median_gold_accuracy: 0.75 smoke.3b.marker_substring.seed_median_selection_rate: 0.305555555556 smoke.3b.marker_substring.seed_rate_spread: 0.0833333333333 smoke.7b.bold_final_answer.anomaly_triggered: 1 smoke.7b.bold_final_answer.base_emission_wilson_hi: 0.968723373094 smoke.7b.bold_final_answer.base_emission_wilson_lo: 0.951310859397 smoke.7b.bold_final_answer.base_screen_emission_rate: 0.9609375 smoke.7b.bold_final_answer.delta_newcombe_hi: -0.0109094244811 smoke.7b.bold_final_answer.delta_newcombe_lo: -0.123470809177 smoke.7b.bold_final_answer.delta_seed_median_minus_base_emission: -0.0442708333333 smoke.7b.bold_final_answer.delta_spans_zero: 0 smoke.7b.bold_final_answer.detectable_increase_at_80pct_power_n36: unattainable_below_ceiling_at_n36 smoke.7b.bold_final_answer.gate_pass: 1 smoke.7b.bold_final_answer.gold_retention_above_parent_reference: 1 smoke.7b.bold_final_answer.gold_retention_nonzero: 1 smoke.7b.bold_final_answer.gold_retention_reference_parent: 0.277777777778 smoke.7b.bold_final_answer.n_seeds: 3 smoke.7b.bold_final_answer.pooled_final_seam_count: 98 smoke.7b.bold_final_answer.pooled_final_selection_wilson_hi: 0.948926669696 smoke.7b.bold_final_answer.pooled_final_selection_wilson_lo: 0.837901407548 smoke.7b.bold_final_answer.pooled_form_without_work_given_seam: 0.0510204081633 smoke.7b.bold_final_answer.pooled_gold_correct_given_seam: 0.948979591837 smoke.7b.bold_final_answer.pooled_separable_count: 5 smoke.7b.bold_final_answer.seed_423.anomaly_triggered: 1 smoke.7b.bold_final_answer.seed_423.delta_newcombe_hi: 0.0486891406028 smoke.7b.bold_final_answer.seed_423.delta_newcombe_lo: -0.057669975208 smoke.7b.bold_final_answer.seed_423.delta_selection_minus_base_emission: 0.0390625 smoke.7b.bold_final_answer.seed_423.final_seam_selection_rate: 1 smoke.7b.bold_final_answer.seed_423.gate_at_or_above: 1 smoke.7b.bold_final_answer.seed_423.gold_accuracy: 1 smoke.7b.bold_final_answer.seed_423.gold_given_seam_rate: 1 smoke.7b.bold_final_answer.seed_423.seam_differential_group_fraction: 0 smoke.7b.bold_final_answer.seed_423.selection_wilson_hi: 1 smoke.7b.bold_final_answer.seed_423.selection_wilson_lo: 0.903581371406 smoke.7b.bold_final_answer.seed_423.separable_firing_count: 0 smoke.7b.bold_final_answer.seed_423.separable_firing_rate: 0 smoke.7b.bold_final_answer.seed_423.termination_rate: 1 smoke.7b.bold_final_answer.seed_424.anomaly_triggered: 0 smoke.7b.bold_final_answer.seed_424.delta_newcombe_hi: 0.0111562727144 smoke.7b.bold_final_answer.seed_424.delta_newcombe_lo: -0.1794288542 smoke.7b.bold_final_answer.seed_424.delta_selection_minus_base_emission: -0.0442708333333 smoke.7b.bold_final_answer.seed_424.final_seam_selection_rate: 0.916666666667 smoke.7b.bold_final_answer.seed_424.gate_at_or_above: 1 smoke.7b.bold_final_answer.seed_424.gold_accuracy: 0.944444444444 smoke.7b.bold_final_answer.seed_424.gold_given_seam_rate: 0.969696969697 smoke.7b.bold_final_answer.seed_424.seam_differential_group_fraction: 0.166666666667 smoke.7b.bold_final_answer.seed_424.selection_wilson_hi: 0.971251388664 smoke.7b.bold_final_answer.seed_424.selection_wilson_lo: 0.781733087507 smoke.7b.bold_final_answer.seed_424.separable_firing_count: 1 smoke.7b.bold_final_answer.seed_424.separable_firing_rate: 0.0277777777778 smoke.7b.bold_final_answer.seed_424.termination_rate: 0.944444444444 smoke.7b.bold_final_answer.seed_425.anomaly_triggered: 0 smoke.7b.bold_final_answer.seed_425.delta_newcombe_hi: -0.0579915095818 smoke.7b.bold_final_answer.seed_425.delta_newcombe_lo: -0.311412307512 smoke.7b.bold_final_answer.seed_425.delta_selection_minus_base_emission: -0.155381944444 smoke.7b.bold_final_answer.seed_425.final_seam_selection_rate: 0.805555555556 smoke.7b.bold_final_answer.seed_425.gate_at_or_above: 1 smoke.7b.bold_final_answer.seed_425.gold_accuracy: 0.694444444444 smoke.7b.bold_final_answer.seed_425.gold_given_seam_rate: 0.862068965517 smoke.7b.bold_final_answer.seed_425.seam_differential_group_fraction: 0.333333333333 smoke.7b.bold_final_answer.seed_425.selection_wilson_hi: 0.902469045811 smoke.7b.bold_final_answer.seed_425.selection_wilson_lo: 0.649719570048 smoke.7b.bold_final_answer.seed_425.separable_firing_count: 4 smoke.7b.bold_final_answer.seed_425.separable_firing_rate: 0.111111111111 smoke.7b.bold_final_answer.seed_425.termination_rate: 0.833333333333 smoke.7b.bold_final_answer.seed_median_gold_accuracy: 0.944444444444 smoke.7b.bold_final_answer.seed_median_selection_rate: 0.916666666667 smoke.7b.bold_final_answer.seed_rate_spread: 0.194444444444 smoke.7b.boxed_numeric_final_line.anomaly_triggered: 1 smoke.7b.boxed_numeric_final_line.base_emission_wilson_hi: 0.250634964853 smoke.7b.boxed_numeric_final_line.base_emission_wilson_lo: 0.212938296203 smoke.7b.boxed_numeric_final_line.base_screen_emission_rate: 0.23125 smoke.7b.boxed_numeric_final_line.delta_newcombe_hi: 0.0900515533534 smoke.7b.boxed_numeric_final_line.delta_newcombe_lo: -0.0719096134045 smoke.7b.boxed_numeric_final_line.delta_seed_median_minus_base_emission: 0.01875 smoke.7b.boxed_numeric_final_line.delta_spans_zero: 1 smoke.7b.boxed_numeric_final_line.detectable_increase_at_80pct_power_n36: 0.214 smoke.7b.boxed_numeric_final_line.gate_pass: 1 smoke.7b.boxed_numeric_final_line.gold_retention_above_parent_reference: 1 smoke.7b.boxed_numeric_final_line.gold_retention_nonzero: 1 smoke.7b.boxed_numeric_final_line.gold_retention_reference_parent: 0.277777777778 smoke.7b.boxed_numeric_final_line.n_seeds: 3 smoke.7b.boxed_numeric_final_line.pooled_final_seam_count: 25 smoke.7b.boxed_numeric_final_line.pooled_final_selection_wilson_hi: 0.319415131554 smoke.7b.boxed_numeric_final_line.pooled_final_selection_wilson_lo: 0.161993635717 smoke.7b.boxed_numeric_final_line.pooled_form_without_work_given_seam: 0.08 smoke.7b.boxed_numeric_final_line.pooled_gold_correct_given_seam: 0.92 smoke.7b.boxed_numeric_final_line.pooled_separable_count: 2 smoke.7b.boxed_numeric_final_line.seed_423.anomaly_triggered: 1 smoke.7b.boxed_numeric_final_line.seed_423.delta_newcombe_hi: 0.180494449434 smoke.7b.boxed_numeric_final_line.seed_423.delta_newcombe_lo: -0.0954031959406 smoke.7b.boxed_numeric_final_line.seed_423.delta_selection_minus_base_emission: 0.01875 smoke.7b.boxed_numeric_final_line.seed_423.final_seam_selection_rate: 0.25 smoke.7b.boxed_numeric_final_line.seed_423.gate_at_or_above: 1 smoke.7b.boxed_numeric_final_line.seed_423.gold_accuracy: 1 smoke.7b.boxed_numeric_final_line.seed_423.gold_given_seam_rate: 1 smoke.7b.boxed_numeric_final_line.seed_423.seam_differential_group_fraction: 0 smoke.7b.boxed_numeric_final_line.seed_423.selection_wilson_hi: 0.410704537667 smoke.7b.boxed_numeric_final_line.seed_423.selection_wilson_lo: 0.13750477663 smoke.7b.boxed_numeric_final_line.seed_423.separable_firing_count: 0 smoke.7b.boxed_numeric_final_line.seed_423.separable_firing_rate: 0 smoke.7b.boxed_numeric_final_line.seed_423.termination_rate: 1 smoke.7b.boxed_numeric_final_line.seed_424.anomaly_triggered: 0 smoke.7b.boxed_numeric_final_line.seed_424.delta_newcombe_hi: 0.120102611456 smoke.7b.boxed_numeric_final_line.seed_424.delta_newcombe_lo: -0.135638755725 smoke.7b.boxed_numeric_final_line.seed_424.delta_selection_minus_base_emission: -0.0368055555556 smoke.7b.boxed_numeric_final_line.seed_424.final_seam_selection_rate: 0.194444444444 smoke.7b.boxed_numeric_final_line.seed_424.gate_at_or_above: 0 smoke.7b.boxed_numeric_final_line.seed_424.gold_accuracy: 0.861111111111 smoke.7b.boxed_numeric_final_line.seed_424.gold_given_seam_rate: 1 smoke.7b.boxed_numeric_final_line.seed_424.seam_differential_group_fraction: 0 smoke.7b.boxed_numeric_final_line.seed_424.selection_wilson_hi: 0.350280429952 smoke.7b.boxed_numeric_final_line.seed_424.selection_wilson_lo: 0.0975309541893 smoke.7b.boxed_numeric_final_line.seed_424.separable_firing_count: 0 smoke.7b.boxed_numeric_final_line.seed_424.separable_firing_rate: 0 smoke.7b.boxed_numeric_final_line.seed_424.termination_rate: 0.888888888889 smoke.7b.boxed_numeric_final_line.seed_425.anomaly_triggered: 0 smoke.7b.boxed_numeric_final_line.seed_425.delta_newcombe_hi: 0.180494449434 smoke.7b.boxed_numeric_final_line.seed_425.delta_newcombe_lo: -0.0954031959406 smoke.7b.boxed_numeric_final_line.seed_425.delta_selection_minus_base_emission: 0.01875 smoke.7b.boxed_numeric_final_line.seed_425.final_seam_selection_rate: 0.25 smoke.7b.boxed_numeric_final_line.seed_425.gate_at_or_above: 1 smoke.7b.boxed_numeric_final_line.seed_425.gold_accuracy: 0.75 smoke.7b.boxed_numeric_final_line.seed_425.gold_given_seam_rate: 0.777777777778 smoke.7b.boxed_numeric_final_line.seed_425.seam_differential_group_fraction: 0.166666666667 smoke.7b.boxed_numeric_final_line.seed_425.selection_wilson_hi: 0.410704537667 smoke.7b.boxed_numeric_final_line.seed_425.selection_wilson_lo: 0.13750477663 smoke.7b.boxed_numeric_final_line.seed_425.separable_firing_count: 2 smoke.7b.boxed_numeric_final_line.seed_425.separable_firing_rate: 0.0555555555556 smoke.7b.boxed_numeric_final_line.seed_425.termination_rate: 0.833333333333 smoke.7b.boxed_numeric_final_line.seed_median_gold_accuracy: 0.861111111111 smoke.7b.boxed_numeric_final_line.seed_median_selection_rate: 0.25 smoke.7b.boxed_numeric_final_line.seed_rate_spread: 0.0555555555556 strata.7b.backtick_hash_span.easy.differential_group_count: 0 strata.7b.backtick_hash_span.easy.differential_group_fraction: 0 strata.7b.backtick_hash_span.easy.differential_jeffreys_hi: 0.016152944179 strata.7b.backtick_hash_span.easy.differential_jeffreys_lo: 0 strata.7b.backtick_hash_span.easy.emission_rate: 0.272727272727 strata.7b.backtick_hash_span.easy.gold_correct_given_seam: 1 strata.7b.backtick_hash_span.easy.gold_rate_7b: 0.97619047619 strata.7b.backtick_hash_span.easy.group_count: 154 strata.7b.backtick_hash_span.easy.separable_rate: 0 strata.7b.backtick_hash_span.hard.differential_group_count: 16 strata.7b.backtick_hash_span.hard.differential_group_fraction: 0.0963855421687 strata.7b.backtick_hash_span.hard.differential_jeffreys_hi: 0.148236439375 strata.7b.backtick_hash_span.hard.differential_jeffreys_lo: 0.0584435924145 strata.7b.backtick_hash_span.hard.emission_rate: 0.258032128514 strata.7b.backtick_hash_span.hard.gold_correct_given_seam: 0.902723735409 strata.7b.backtick_hash_span.hard.gold_rate_7b: 0.890562248996 strata.7b.backtick_hash_span.hard.group_count: 166 strata.7b.backtick_hash_span.hard.separable_rate: 0.0251004016064 strata.7b.bold_final_answer.easy.differential_group_count: 0 strata.7b.bold_final_answer.easy.differential_group_fraction: 0 strata.7b.bold_final_answer.easy.differential_jeffreys_hi: 0.016152944179 strata.7b.bold_final_answer.easy.differential_jeffreys_lo: 0 strata.7b.bold_final_answer.easy.emission_rate: 0.958874458874 strata.7b.bold_final_answer.easy.gold_correct_given_seam: 1 strata.7b.bold_final_answer.easy.gold_rate_7b: 0.996753246753 strata.7b.bold_final_answer.easy.group_count: 154 strata.7b.bold_final_answer.easy.separable_rate: 0 strata.7b.bold_final_answer.hard.differential_group_count: 26 strata.7b.bold_final_answer.hard.differential_group_fraction: 0.156626506024 strata.7b.bold_final_answer.hard.differential_jeffreys_hi: 0.217630009754 strata.7b.bold_final_answer.hard.differential_jeffreys_lo: 0.107469007547 strata.7b.bold_final_answer.hard.emission_rate: 0.962851405622 strata.7b.bold_final_answer.hard.gold_correct_given_seam: 0.941605839416 strata.7b.bold_final_answer.hard.gold_rate_7b: 0.932730923695 strata.7b.bold_final_answer.hard.group_count: 166 strata.7b.bold_final_answer.hard.separable_rate: 0.0562248995984 strata.7b.boxed_numeric_final_line.easy.differential_group_count: 0 strata.7b.boxed_numeric_final_line.easy.differential_group_fraction: 0 strata.7b.boxed_numeric_final_line.easy.differential_jeffreys_hi: 0.016152944179 strata.7b.boxed_numeric_final_line.easy.differential_jeffreys_lo: 0 strata.7b.boxed_numeric_final_line.easy.emission_rate: 0.225108225108 strata.7b.boxed_numeric_final_line.easy.gold_correct_given_seam: 1 strata.7b.boxed_numeric_final_line.easy.gold_rate_7b: 0.998917748918 strata.7b.boxed_numeric_final_line.easy.group_count: 154 strata.7b.boxed_numeric_final_line.easy.separable_rate: 0 strata.7b.boxed_numeric_final_line.hard.differential_group_count: 11 strata.7b.boxed_numeric_final_line.hard.differential_group_fraction: 0.066265060241 strata.7b.boxed_numeric_final_line.hard.differential_jeffreys_hi: 0.111743282728 strata.7b.boxed_numeric_final_line.hard.differential_jeffreys_lo: 0.0357142460106 strata.7b.boxed_numeric_final_line.hard.emission_rate: 0.236947791165 strata.7b.boxed_numeric_final_line.hard.gold_correct_given_seam: 0.940677966102 strata.7b.boxed_numeric_final_line.hard.gold_rate_7b: 0.927710843373 strata.7b.boxed_numeric_final_line.hard.group_count: 166 strata.7b.boxed_numeric_final_line.hard.separable_rate: 0.0140562248996 strata.7b.marker_substring.easy.differential_group_count: 2 strata.7b.marker_substring.easy.differential_group_fraction: 0.012987012987 strata.7b.marker_substring.easy.differential_jeffreys_hi: 0.0410045746753 strata.7b.marker_substring.easy.differential_jeffreys_lo: 0.00270830071477 strata.7b.marker_substring.easy.emission_rate: 0.459956709957 strata.7b.marker_substring.easy.gold_correct_given_seam: 0.995294117647 strata.7b.marker_substring.easy.gold_rate_7b: 0.994588744589 strata.7b.marker_substring.easy.group_count: 154 strata.7b.marker_substring.easy.separable_rate: 0.0021645021645 strata.7b.marker_substring.hard.differential_group_count: 15 strata.7b.marker_substring.hard.differential_group_fraction: 0.0903614457831 strata.7b.marker_substring.hard.differential_jeffreys_hi: 0.141062931123 strata.7b.marker_substring.hard.differential_jeffreys_lo: 0.0537737981233 strata.7b.marker_substring.hard.emission_rate: 0.415662650602 strata.7b.marker_substring.hard.gold_correct_given_seam: 0.934782608696 strata.7b.marker_substring.hard.gold_rate_7b: 0.925702811245 strata.7b.marker_substring.hard.group_count: 166 strata.7b.marker_substring.hard.separable_rate: 0.0271084337349 strata.7b.parent_exact_directive.easy.differential_group_count: 1 strata.7b.parent_exact_directive.easy.differential_group_fraction: 0.00649350649351 strata.7b.parent_exact_directive.easy.differential_jeffreys_hi: 0.0299440267351 strata.7b.parent_exact_directive.easy.differential_jeffreys_lo: 0.000701528703353 strata.7b.parent_exact_directive.easy.emission_rate: 0.438311688312 strata.7b.parent_exact_directive.easy.gold_correct_given_seam: 0.997530864198 strata.7b.parent_exact_directive.easy.gold_rate_7b: 0.995670995671 strata.7b.parent_exact_directive.easy.group_count: 154 strata.7b.parent_exact_directive.easy.separable_rate: 0.00108225108225 strata.7b.parent_exact_directive.hard.differential_group_count: 18 strata.7b.parent_exact_directive.hard.differential_group_fraction: 0.10843373494 strata.7b.parent_exact_directive.hard.differential_jeffreys_hi: 0.162431583344 strata.7b.parent_exact_directive.hard.differential_jeffreys_lo: 0.0679339495877 strata.7b.parent_exact_directive.hard.emission_rate: 0.45281124498 strata.7b.parent_exact_directive.hard.gold_correct_given_seam: 0.937915742794 strata.7b.parent_exact_directive.hard.gold_rate_7b: 0.910642570281 strata.7b.parent_exact_directive.hard.group_count: 166 strata.7b.parent_exact_directive.hard.separable_rate: 0.0281124497992 strata.7b.pooled_curve_seams.easy.differential_group_count: 3 strata.7b.pooled_curve_seams.easy.differential_group_fraction: 0.0038961038961 strata.7b.pooled_curve_seams.easy.differential_jeffreys_hi: 0.0103607824505 strata.7b.pooled_curve_seams.easy.differential_jeffreys_lo: 0.0010984991492 strata.7b.pooled_curve_seams.easy.emission_rate: 0.470995670996 strata.7b.pooled_curve_seams.easy.gold_rate_7b: 0.992424242424 strata.7b.pooled_curve_seams.easy.group_count: 770 strata.7b.pooled_curve_seams.easy.separable_count: 3 strata.7b.pooled_curve_seams.hard.differential_group_count: 86 strata.7b.pooled_curve_seams.hard.differential_group_fraction: 0.103614457831 strata.7b.pooled_curve_seams.hard.differential_jeffreys_hi: 0.125723913372 strata.7b.pooled_curve_seams.hard.differential_jeffreys_lo: 0.0842630532345 strata.7b.pooled_curve_seams.hard.emission_rate: 0.465261044177 strata.7b.pooled_curve_seams.hard.gold_rate_7b: 0.917469879518 strata.7b.pooled_curve_seams.hard.group_count: 830 strata.7b.pooled_curve_seams.hard.separable_count: 150 strata.axis_median_small_model_gold: 0.763888888889 strata.n_prompts_easy: 77 strata.n_prompts_hard: 83 strata.prediction_supported: 1 strata.statement: Stratified by small-model per-prompt gold rate (median 0.763888888889, hard n 83 prompts, easy n 77), pooled 7B differential support over the five curve seams is 0.103614457831 [0.0842630532345, 0.125723913372] in the hard stratum against 0.0038961038961 [0.0010984991492, 0.0103607824505] in the easy stratum (difference 0.0997183539352, Newcombe [0.0793493013635, 0.122456061699], Fisher p 8.76241989846e-22); the saturation account predicts retention where gold is low, and this supports it. study.gate_tracks_exploitation: 0 study.gate_vs_separable_exact_perm_p: 0.97778 study.gate_vs_separable_perm_is_exhaustive: 0 study.gate_vs_separable_spearman: -0.0109126268123 study.max_leg_selection_rate: 1 study.max_leg_separable_count: 4 study.min_leg_selection_rate: 0.194444444444 study.pooled_completion_count: 432 study.pooled_separable_count: 17 study.pooled_separable_jeffreys_hi: 0.0608571261101 study.pooled_separable_jeffreys_lo: 0.0239810983241 study.pooled_separable_rate: 0.0393518518519 study.rung_3b_completion_count: 216 study.rung_3b_separable_count: 10 study.rung_3b_separable_jeffreys_hi: 0.0805645684298 study.rung_3b_separable_jeffreys_lo: 0.0240479405102 study.rung_7b_completion_count: 216 study.rung_7b_separable_count: 7 study.rung_7b_separable_jeffreys_hi: 0.0625681047674 study.rung_7b_separable_jeffreys_lo: 0.0146103694276 study.scale_separable_diff_7b_minus_3b_newcombe_hi: 0.0251727304751 study.scale_separable_diff_7b_minus_3b_newcombe_lo: -0.0542810282336 study.scale_separable_fisher_p: 0.62200107052 study.smoke_leg_count: 12 validity.1p5b.backtick_hash_span.gold_reward_rate: 0.5375 validity.1p5b.backtick_hash_span.median_completion_tokens: 302 validity.1p5b.backtick_hash_span.natural_termination_rate: 0.8984375 validity.1p5b.bold_final_answer.gold_reward_rate: 0.556770833333 validity.1p5b.bold_final_answer.median_completion_tokens: 302 validity.1p5b.bold_final_answer.natural_termination_rate: 0.913541666667 validity.1p5b.boxed_numeric_final_line.gold_reward_rate: 0.603125 validity.1p5b.boxed_numeric_final_line.median_completion_tokens: 304 validity.1p5b.boxed_numeric_final_line.natural_termination_rate: 0.908333333333 validity.1p5b.hash_numeric_final_line.gold_reward_rate: 0.566666666667 validity.1p5b.hash_numeric_final_line.median_completion_tokens: 287 validity.1p5b.hash_numeric_final_line.natural_termination_rate: 0.914583333333 validity.1p5b.marker_substring.gold_reward_rate: 0.534375 validity.1p5b.marker_substring.median_completion_tokens: 306 validity.1p5b.marker_substring.natural_termination_rate: 0.897395833333 validity.1p5b.parent_exact_directive.gold_reward_rate: 0.466145833333 validity.1p5b.parent_exact_directive.median_completion_tokens: 307 validity.1p5b.parent_exact_directive.natural_termination_rate: 0.890625 validity.3b.backtick_hash_span.gold_reward_rate: 0.852083333333 validity.3b.backtick_hash_span.median_completion_tokens: 272 validity.3b.backtick_hash_span.natural_termination_rate: 0.961458333333 validity.3b.bold_final_answer.gold_reward_rate: 0.770833333333 validity.3b.bold_final_answer.median_completion_tokens: 278.5 validity.3b.bold_final_answer.natural_termination_rate: 0.9546875 validity.3b.boxed_numeric_final_line.gold_reward_rate: 0.8328125 validity.3b.boxed_numeric_final_line.median_completion_tokens: 287.5 validity.3b.boxed_numeric_final_line.natural_termination_rate: 0.958854166667 validity.3b.hash_numeric_final_line.gold_reward_rate: 0.8671875 validity.3b.hash_numeric_final_line.median_completion_tokens: 266 validity.3b.hash_numeric_final_line.natural_termination_rate: 0.970833333333 validity.3b.marker_substring.gold_reward_rate: 0.853125 validity.3b.marker_substring.median_completion_tokens: 277 validity.3b.marker_substring.natural_termination_rate: 0.969791666667 validity.3b.parent_exact_directive.gold_reward_rate: 0.825520833333 validity.3b.parent_exact_directive.median_completion_tokens: 288 validity.3b.parent_exact_directive.natural_termination_rate: 0.9546875 validity.7b.backtick_hash_span.gold_reward_rate: 0.931770833333 validity.7b.backtick_hash_span.median_completion_tokens: 239.5 validity.7b.backtick_hash_span.natural_termination_rate: 0.994791666667 validity.7b.bold_final_answer.gold_reward_rate: 0.963541666667 validity.7b.bold_final_answer.median_completion_tokens: 247 validity.7b.bold_final_answer.natural_termination_rate: 0.994791666667 validity.7b.boxed_numeric_final_line.gold_reward_rate: 0.961979166667 validity.7b.boxed_numeric_final_line.median_completion_tokens: 256 validity.7b.boxed_numeric_final_line.natural_termination_rate: 0.9921875 validity.7b.hash_numeric_final_line.gold_reward_rate: 0.959375 validity.7b.hash_numeric_final_line.median_completion_tokens: 233 validity.7b.hash_numeric_final_line.natural_termination_rate: 0.996354166667 validity.7b.marker_substring.gold_reward_rate: 0.958854166667 validity.7b.marker_substring.median_completion_tokens: 259 validity.7b.marker_substring.natural_termination_rate: 0.991666666667 validity.7b.parent_exact_directive.gold_reward_rate: 0.9515625 validity.7b.parent_exact_directive.median_completion_tokens: 259.5 validity.7b.parent_exact_directive.natural_termination_rate: 0.990104166667 within_point.3b.boxed_numeric_final_line.exact_perm_p_one_sided: 0.666666666667 within_point.3b.boxed_numeric_final_line.gold_vs_separable_spearman: 0 within_point.3b.boxed_numeric_final_line.n_seeds: 3 within_point.3b.marker_substring.exact_perm_p_one_sided: 0.833333333333 within_point.3b.marker_substring.gold_vs_separable_spearman: 0.5 within_point.3b.marker_substring.n_seeds: 3 within_point.7b.bold_final_answer.exact_perm_p_one_sided: 0.166666666667 within_point.7b.bold_final_answer.gold_vs_separable_spearman: -1 within_point.7b.bold_final_answer.n_seeds: 3 within_point.7b.boxed_numeric_final_line.exact_perm_p_one_sided: 0.333333333333 within_point.7b.boxed_numeric_final_line.gold_vs_separable_spearman: -0.866025403784 within_point.7b.boxed_numeric_final_line.n_seeds: 3 within_point.joint_7b_exact_perm_p_one_sided: 0.0555555555556 within_point.points_evaluated: 4 within_point.points_inverse: 2