{"id":"c9200b56-383b-4cd4-8a2a-bd2409a953cd","arxiv_id":"2611.26510","paper_version":1,"verdict":"CONDITIONAL","confidence":"MODERATE","novelty_score":6.0,"correctness_risk":"medium","formal_verification":"none","parameter_count":5,"one_line_summary":"Guiding teacher rollouts with natural-language skills and keeping those skills in the SFT input raises mean student performance over unguided-trajectory SFT by 1.42 to 8.99 points, with 2 of 6 paired confidence intervals including zero.","lead":"This paper asks whether a short written 'skill' shown to an AI teacher before it works makes its demonstrations train a better student, and reports it usually does. An autonomous agent writes, tests, and refines those skills, lifting mean student scores by 1.42 to 8.99 points over standard training across six benchmark-model combinations.","discovery_kind":"new_method","skeptic_critique":{"audit":{"quote":"At the end of autonomous experimentation, \\papername{} selects\na final checkpoint based on accumulated evaluation feedback\nand records its choice in a final report.","status":"reviewer_error","question":null,"consequence":null,"objection_id":"O1","published_as":"withdrawn"},"model":"deepseek/deepseek-v4-flash","headline":"Withdrawn: Pith's reviewer misread the paper.","rationale":"","tokens_in":26351,"tokens_out":803,"duration_ms":7508,"concrete_test":"","verdict_should_be":null,"load_bearing_attack":"The paper's words: “At the end of autonomous experimentation, \\papername{} selects\na final checkpoint based on accumulated evaluation feedback\nand records its choice in a final report.”","agreement_with_reader":null},"referee_report":{"model":"deepseek/deepseek-v4-flash","summary":"The paper proposes Skill-FT (Skill-Guided Fine-Tuning) and AutoMaster, an autonomous post-training system where an LLM agent designs natural-language skills to shape teacher rollouts, selects demonstrations, and iteratively refines interventions across rounds. The central claim is that behavioral shaping via skills makes a fixed teacher model more effective, improving student performance over Vanilla SFT on Terminal, Search, and SWE benchmarks with 4B and 9B models. The paper also reports that retaining skill instructions during SFT (even though they are absent at evaluation) yields higher performance than removing them.","tokens_in":26632,"tokens_out":2377,"duration_ms":16709,"significance":"Not published. Pith's referee wrote this assessment before its objections were checked against the paper, and 3 of its 3 major objections did not survive that check.","major_comments":[],"minor_comments":[{"quote":"Results are mean $\\pm$ SD over three evaluation runs per checkpoint.","comment":"The table reports mean ± SD over three evaluation runs per checkpoint. For Vanilla SFT, the SD is sometimes larger than for AutoMaster (e.g., Doc2Repo 4B: 4.00 vs 3.58). The paper should clarify whether the three runs are independent (different random seeds) or the same checkpoint evaluated three times. If the latter, the SD reflects evaluation noise, not training variability, and the comparison may be affected.","section":"§4.1 (Main Results), Table 1","consequence":null,"audit_status":null,"objection_id":null,"quote_matched":true,"required_revision":null,"authors_acknowledge":null,"authors_acknowledge_p":null},{"quote":"At the end of autonomous experimentation, \\papername{} selects\na final checkpoint based on accumulated evaluation feedback\nand records its choice in a final report.","comment":"The paper states that at the end of autonomous experimentation, AutoMaster selects a final checkpoint based on accumulated evaluation feedback and records its choice. The selection criteria are not specified. The paper should describe how the agent chooses the final checkpoint (e.g., highest mean, highest single run, or some other heuristic) to allow reproducibility.","section":"§4.2 (Model Evaluation), Appendix A.4","consequence":null,"audit_status":null,"objection_id":null,"quote_matched":true,"required_revision":null,"authors_acknowledge":null,"authors_acknowledge_p":null},{"quote":"To test whether the choice of training tasks alone explains the gains,\nwe construct \\textbf{Task-Matched SFT} from unguided rollouts by the\nsame teacher, matching both the task IDs and the per-task sample counts\nof \\papername{}'s final training set.","comment":"The Task-Matched SFT baseline uses unguided rollouts matched to AutoMaster's final training tasks, but the paper does not specify whether the unguided rollouts are from the same teacher or from a different run. The caption says 'unguided rollouts from the same teacher', which is clear, but the text should also state that the unguided rollouts are from the initial unguided teacher rollouts (not from a later round).","section":"§4.3 (Ablation Studies), Figure 3","consequence":null,"audit_status":null,"objection_id":null,"quote_matched":true,"required_revision":null,"authors_acknowledge":null,"authors_acknowledge_p":null},{"quote":"Across all six settings, \\textbf{Keep} exceeds \\textbf{Remove} in mean score by 2.15--4.01 percentage points (Figure~6).","comment":"The comparison between Keep and Remove is a within-trajectory ablation, but the paper does not report whether the difference is statistically significant. Given the small number of evaluation runs (three), the paper should provide confidence intervals or a significance test for the Keep vs Remove comparison.","section":"§5.2 (Preserving Skill Context), Figure 6","consequence":null,"audit_status":null,"objection_id":null,"quote_matched":true,"required_revision":null,"authors_acknowledge":null,"authors_acknowledge_p":null},{"quote":"For \\papername{}, the agent starts with trajectories generated\nby the initial target model and the unguided teacher on the\ntraining task pool, with a budget of five outer-loop rounds.","comment":"The paper shows performance across five rounds, but does not report how many rounds the agent actually ran in each setting (the budget is five, but the agent may stop earlier). The paper should clarify whether all five rounds were completed for each setting and whether the final checkpoint is always from the round with the highest mean.","section":"§5.3 (Autonomous Experimentation), Figure 7","consequence":null,"audit_status":null,"objection_id":null,"quote_matched":true,"required_revision":null,"authors_acknowledge":null,"authors_acknowledge_p":null},{"quote":"Prompt and skill optimization methods use execution feedback to refine natural-language guidance without updating model parameters~\\citep{agrawal2025gepa,shi2026skillkd,ge2026skillgym}.","comment":"The related work section cites several papers but does not discuss the key difference between AutoMaster and prior work on prompt optimization (e.g., GEPA) or skill libraries (e.g., SkillRL). The paper should clarify that AutoMaster uses skills to shape teacher rollouts for SFT data generation, rather than optimizing prompts for the student's own performance.","section":"§2 (Related Work)","consequence":null,"audit_status":null,"objection_id":null,"quote_matched":true,"required_revision":null,"authors_acknowledge":null,"authors_acknowledge_p":null},{"quote":null,"comment":"The paper reports gains of 1.42-8.99 percentage points, but the baseline Vanilla SFT uses 512 trajectories randomly sampled from successful unguided teacher rollouts. The paper should discuss whether the random sampling of 512 from a larger pool introduces variance that could affect the comparison, and whether AutoMaster's selection of 512 trajectories from a potentially different distribution (including some unsuccessful ones?) is a confounding factor.","section":"§4.1 (Main Results), Table 1","consequence":null,"audit_status":null,"objection_id":null,"quote_matched":false,"required_revision":null,"authors_acknowledge":null,"authors_acknowledge_p":null},{"quote":"For Terminal and Search, we compare binary task outcomes.\nAcross these four corpora, 8.79--20.70\\% of the pairs contain\na successful \\papername{} demonstration and an unsuccessful\nunguided rollout.\nThe remaining pairs succeed in both conditions.","comment":"Table A5 shows that for Terminal and Search, 0% of pairs have 'Success → failure' (i.e., the unguided rollout succeeds but the AutoMaster demonstration fails). This suggests that skill guidance rarely hurts task success. However, the paper should note that this is on the training tasks, not the evaluation set, and that the agent may avoid applying skills to tasks where they would hurt.","section":"§4.3 (Ablation Studies), Table A5","consequence":null,"audit_status":null,"objection_id":null,"quote_matched":true,"required_revision":null,"authors_acknowledge":null,"authors_acknowledge_p":null},{"quote":"To examine the behaviors targeted by \\method{}, Figure~\\ref{fig:behavioral_shaping_cases} contrasts unguided and skill-guided teacher trajectories on the same tasks across three domains.","comment":"The behavioral shaping examples are illustrative but the paper does not quantify how often such improvements occur across the training set. The paper should provide a systematic analysis (e.g., human evaluation or automated metrics) of how often skill guidance changes teacher behavior in the intended direction.","section":"§5.1 (Behavioral Shaping), Figure 5","consequence":null,"audit_status":null,"objection_id":null,"quote_matched":true,"required_revision":null,"authors_acknowledge":null,"authors_acknowledge_p":null},{"quote":"Across all domains, we use a temperature of 1.0, a top-$p$ of 0.95,\na maximum of 65,500 tokens per model response, and a limit of\n600 execution steps per task.","comment":"The paper states that evaluation uses a temperature of 1.0 and top-p 0.95. The paper should clarify whether these settings are the same as those used during teacher rollouts (temperature 1.0, top-p 1.0) and whether the difference in top-p could affect the comparison.","section":"§4.2 (Model Evaluation), Appendix A.4","consequence":null,"audit_status":null,"objection_id":null,"quote_matched":true,"required_revision":null,"authors_acknowledge":null,"authors_acknowledge_p":null}],"open_questions":null,"recommendation":"uncertain"},"author_rebuttal":null,"desk_editor":{"model":"deepseek/deepseek-v4-flash","letter":"The paper's core claim is that natural-language skills appended to the teacher's system prompt can shape which behaviors appear in successful trajectories, and that these shaped demonstrations train better students than unguided ones, all with the teacher, task pool, training-set size, and SFT recipe held fixed. That claim is mostly supported by the evidence, and the authors report it honestly: paired bootstrap confidence intervals, a task-matched SFT control, and component ablations are all good practice, and they disclose that the 4B gains on BrowseComp and Doc2Repo have intervals including zero.\n\nThe cleanest result is the Keep-versus-Remove comparison: identical trajectories and SFT recipe, skill instructions either retained or dropped from the training input, no skills at evaluation, and Keep beats Remove by 2.15 to 4.01 points in all six settings. That is a controlled internalization finding and, as far as I can tell from the cited literature, genuinely new. It deserves to be the centerpiece.\n\nTwo soft spots, in proportion. First, the Appendix's Model Evaluation section states that the agent selects the final AutoMaster checkpoint based on accumulated evaluation scores and trajectories over up to five rounds, then re-evaluates it three times, while Vanilla SFT is a single fixed run. The later rounds section shows per-round variation of several points, so part of the reported headroom may be best-of-rounds selection rather than skill-guided generation alone. This is the main thing a referee should push on. The component ablations do not fully resolve it, since those variants also get agent selection. Second, the two CIs already including zero mean the effect is not established for 4B BrowseComp or 4B Doc2Repo even before that confound.\n\nMinor notes: the closest skill-distillation baselines, Skill-SD and Skill-to-LoRA, are cited but not run, and several baselines are 2026 arXiv preprints, which made related-work positioning harder for me to check. One claimed result I could not verify: the nine-point stencil claim in the Terminal case relies on a figure I could not see.\n\nThis paper deserves a serious referee. The internalization result is well controlled, the main comparison is carefully reported, and the selection confound is fixable. A revision that either holds out an evaluation-excluded selection criterion or analyzes the final round's checkpoint under the same protocol as Vanilla SFT would substantially firm up the central claim.","headline":"Skill-guided teacher rollouts plus retaining skill context during SFT give consistent student gains over Vanilla SFT in 4 of 6 settings, with the headline caveat that the agent picks the final checkpoint from evaluation feedback across up to five rounds.","tokens_in":26900,"tokens_out":872,"duration_ms":6857},"pith_extraction":{"msc":[],"pacs":[],"model":"deepseek/deepseek-v4-flash","headline":"The paper argues that behavioral shaping via natural-language skills can make a fixed teacher model more effective for post-training, and that retaining skill instructions during SFT improves student performance even when skills are…","keywords":["post-training","supervised fine-tuning","skill-guided fine-tuning","LLM agents","behavioral shaping","autonomous post-training","teacher-student distillation","natural-language skills"],"falsifier":"Run AutoMaster with the same teacher, task pool, and SFT recipe but with the agent's skill design and data selection replaced by random selection of 512 trajectories from the same pool of unguided successful rollouts, and compare to Vanilla SFT on a single fixed run. If the gains disappear, the headroom is due to selection rather than behavioral shaping.","tokens_in":28066,"feed_emoji":"🎓","tokens_out":1380,"duration_ms":11651,"temperature":0.2,"pith_summary":"The paper proposes Skill-FT (Skill-Guided Fine-Tuning), a method that uses natural-language skills to shape the behaviors a teacher model demonstrates during rollout, then fine-tunes a student on the resulting trajectories. The authors introduce AutoMaster, an autonomous system where an LLM agent designs, tests, and refines these skills through iterative rounds of trajectory analysis, skill design, data selection, and evaluation feedback. Across Terminal, Search, and software engineering tasks with 4B and 9B student models, AutoMaster improves mean performance over standard SFT on successful unguided teacher trajectories by 1.42–8.99 percentage points, with the teacher, task pool, training set size, and SFT recipe held fixed. The paper further reports that retaining skill instructions in the SFT input yields higher mean performance than removing them across all six settings, even though students are evaluated without skills. The authors argue that behavioral shaping can make a fixed model a more effective teacher, and that preserving skill context during training supports better behavioral internalization.","feed_headline":"Authors report behavioral shaping via skills improves post-training from a fixed teacher","feed_subtitle":"AutoMaster uses natural-language skills to guide teacher rollouts, boosting student performance by 1.42–8.99 percentage points over…","key_machinery":"Skill-FT (Skill-Guided Fine-Tuning): natural-language skills appended to the teacher's system prompt to suppress undesirable habits and elicit effective behaviors during rollout. AutoMaster: an autonomous system with an inner loop (skill design, pilot rollouts, refinement, data selection) and an outer loop (SFT from a fixed initial checkpoint, evaluation feedback, iterative refinement). The central mechanism is behavioral shaping of teacher demonstrations before student training, combined with retaining skill instructions in the training input.","core_discovery":"The paper claims that a fixed teacher model can be made more effective for post-training by shaping its behavioral patterns during rollout using natural-language skills, and that the resulting skill-guided demonstrations improve student performance beyond what is achievable by selecting from unguided successful trajectories alone. The authors further claim that retaining the skill instructions in the SFT input, even though they are absent at evaluation, improves student performance compared to removing them, suggesting that preserving the context behind demonstrated reasoning and actions aids behavioral internalization.","headline_check":{"rule":"report_a_claim_v1","sent_back":false,"unattributed":{}},"pith_inferences":["The approach could be extended to other forms of behavioral guidance beyond natural-language skills, such as structured constraints or reward signals during rollout.","The finding that skill instructions aid internalization even when absent at evaluation raises the question of whether similar context-preserving strategies could benefit other forms of instruction-tuned or demonstration-based learning.","The reliance on a strong LLM agent (GPT-6 Astra) for skill design and refinement means the method's effectiveness may be sensitive to the agent's capabilities; testing with weaker agents would clarify the generality.","The paired confidence intervals for 4B BrowseComp and Doc2Repo include zero, so the claim of improvement in those settings is not statistically established even before considering the selection-bias confound."],"forward_implications":["If correct, the approach suggests that post-training headroom from a fixed teacher is larger than what is captured by distilling successful unguided trajectories alone.","The finding that retaining skill instructions during SFT improves performance, despite a training-evaluation prompt mismatch, challenges the assumption that minimizing prompt mismatch is always beneficial.","The results imply that autonomous post-training systems can improve student performance by jointly optimizing how demonstrations are generated and selected, rather than only selecting from a fixed pool.","The larger gains for the 9B model compared to the 4B model suggest that greater student capacity may better leverage skill-guided demonstrations."],"supporting_citations":[{"why":"Establishes the baseline approach of SFT on successful agent trajectories, which AutoMaster compares against.","marker":"chen2023fireact"},{"why":"Provides a related method for recovering demonstrations from teacher failures, which AutoMaster complements by shaping behaviors on successful tasks.","marker":"ta2026proofgen"},{"why":"Selects trajectories exhibiting beneficial behaviors, a related approach that AutoMaster extends by also shaping behaviors during generation.","marker":"jin2026beneficial"},{"why":"Trains skill-specific adapters on skill-guided demonstrations, a related paradigm that AutoMaster compares with by retaining skills in the SFT input.","marker":"zhang2026skilltolora"},{"why":"Uses a skill-conditioned teacher for on-policy self-distillation, a related method that AutoMaster contrasts by focusing on teacher-guided demonstrations.","marker":"wang2026skillsd"},{"why":"Provides the Terminal-Bench 2.1 evaluation benchmark used in the experiments.","marker":"merrill2026terminalbench"},{"why":"Provides the BrowseComp benchmark used for search task evaluation.","marker":"wei2025browsecomp"},{"why":"Provides the Doc2Repo benchmark and evaluation harness for software engineering tasks.","marker":"chen2026beyondswe"}],"cache_read_input_tokens":0,"weakest_assumption_plain":"The reported AutoMaster checkpoint is selected by the agent using accumulated evaluation scores and trajectories over up to five rounds, then re-evaluated three times, while Vanilla SFT is a single fixed run; if the headroom comes from best-of-rounds selection rather than from skill-guided generation and data selection, the central claim collapses to selection bias.","cache_creation_input_tokens":0},"created_at":"2026-11-20T06:26:12.936089+00:00","model_set":{"reader":"deepseek/deepseek-v4-flash"},"pipeline_version":"pith-pipeline@v1.4.1-daily-deepseek-or-medium+deepseek/deepseek-v4-flash+p.d60f54226f00","input_kind":"tex","input_chars":88019,"input_total_chars":88019,"passes_missing":[],"published_standing":"established_narrower","published_confidence":"moderate","objection_audit":{"version":"pith-objection-audit@1","counts":{"raised":4,"confirmed":0,"questions":0,"withdrawn":4,"moved_to_minor":0,"math_error_not_certified":0,"withdrawn_reviewer_error":4,"typesetting_minor_dropped":0,"withdrawn_already_addressed":0,"referee_assessment_unpublished":1},"open_questions":[],"standing":{"basis":["first_pass_verdict"],"standing":"established_narrower","confidence":"moderate"},"standing_rule":"1. No standing is published (not_assessed) when the review read only the abstract, when the referee pass or the objection check did not run, or when the first pass gave no verdict. 2. The first pass's verdict sets the start: ACCEPT is established, CONDITIONAL is established_narrower, REJECT is not_supported. 3. Each objection that survived the check against the paper's text caps the standing: one that breaks the central claim caps it at not_supported, one that narrows it at established_narrower, one that weakens it at supported_with_gaps. The standing is the lowest of the start and the caps. 4. A first-pass REJECT stands only if a surviving objection breaks the central claim. Without one, the standing is the lowest cap of the surviving objections, or not_assessed when none survived. 5. The confidence is the lowest of the first pass's confidence and the confidence of each objection that set the standing, and is no higher than moderate while a question is open or part of the paper's text was not supplied. 6. If any objection was withdrawn because Pith's reviewer misread the paper, the confidence is no higher than moderate, whatever the first pass said: a review that misread the paper once may have misread it elsewhere.","input_kind":"tex","nothing_to_audit":false,"bar":{"enabled":false}},"run_record":{"models":["deepseek/deepseek-v4-flash"],"passes":{"reader":{"model":"deepseek/deepseek-v4-flash","effort":"medium","cost_usd":0.007662,"tokens_in":30948,"tokens_out":14348,"duration_ms":71593,"input_sha256":"f87502c5e4e6145e4d82eaf4f9558af90348ef355b94105a49c1ff38947ed5b4","prompt_sha256":"f9ead730038216984cfb56d7cb34332bdf691828296a1de4245f3d6667d4bf56","message_sha256":"d492989513fa10955a6499e600454ecfeb485cf0041580534396935c740c62ea","cache_read_input_tokens":0,"cache_creation_input_tokens":0},"referee":{"model":"deepseek/deepseek-v4-flash","effort":"medium","cost_usd":0.002856,"tokens_in":26632,"tokens_out":2377,"duration_ms":16709,"input_sha256":"f87502c5e4e6145e4d82eaf4f9558af90348ef355b94105a49c1ff38947ed5b4","prompt_sha256":"a852d99aea0ddfd900e1483f76e20777aae3afde9c016d5b8d6e5db01064c728","message_sha256":"709c7a9c06e4639968195cb63dbd2b0edc186f37e0f954aa549b7c4d5f9d9803","cache_read_input_tokens":0,"cache_creation_input_tokens":0},"skeptic":{"model":"deepseek/deepseek-v4-flash","effort":"medium","cost_usd":0.002544,"tokens_in":26351,"tokens_out":803,"duration_ms":7508,"input_sha256":"f87502c5e4e6145e4d82eaf4f9558af90348ef355b94105a49c1ff38947ed5b4","prompt_sha256":"8f88c0f61488b1e764c317998cddf35fffc33a1f20a3b1be54330d7843386fc7","message_sha256":"68ca3b09ac01a8d5c02cb4a8b0b25a8c27d0c3a81b860d65b1d767d6bfbe9e44","cache_read_input_tokens":0,"cache_creation_input_tokens":0},"circularity":{"model":"deepseek/deepseek-v4-flash","effort":"medium","cost_usd":0.004002,"tokens_in":56946,"tokens_out":4065,"duration_ms":27730,"input_sha256":"f87502c5e4e6145e4d82eaf4f9558af90348ef355b94105a49c1ff38947ed5b4","prompt_sha256":"e9946e39d67b76cc60476379bc32abdc44665b5f0c438659796547bb1041a01f","message_sha256":"426fc0e9e75f574f856a08d66508f78307c23a13f5989905644fcb274e536d4d","cache_read_input_tokens":0,"cache_creation_input_tokens":0},"desk_editor":{"model":"deepseek/deepseek-v4-flash","effort":"medium","cost_usd":0.003895,"tokens_in":26900,"tokens_out":872,"duration_ms":6857,"input_sha256":"f87502c5e4e6145e4d82eaf4f9558af90348ef355b94105a49c1ff38947ed5b4","prompt_sha256":"e7211f6a3d2734c1025ce7e7e6d48f78b0944d0a32d5021f81be63b27721420a","message_sha256":"0ea32d68dd6256626ff4cfac50ba99cac60a22eafd1756f39fbffeaf6af9955c","cache_read_input_tokens":0,"cache_creation_input_tokens":0},"objection_audit":{"model":"deepseek/deepseek-v4-flash","effort":"medium","cost_usd":0.003903,"tokens_in":55101,"tokens_out":4032,"duration_ms":23958,"input_sha256":"f87502c5e4e6145e4d82eaf4f9558af90348ef355b94105a49c1ff38947ed5b4","prompt_sha256":"6c7a6883d5517cf9e4945d83947200a36dcd152b8b56c32baf7500d8b3841d0a","message_sha256":"867fe2ccd5d0d9d5783686ef30eb95e102465a8a42d2a9f918a0fd8b11042e54","cache_read_input_tokens":0,"cache_creation_input_tokens":0},"pith_extraction":{"model":"deepseek/deepseek-v4-flash","effort":"medium","cost_usd":0.002805,"tokens_in":28066,"tokens_out":1380,"duration_ms":11651,"temperature":0.2,"input_sha256":"f87502c5e4e6145e4d82eaf4f9558af90348ef355b94105a49c1ff38947ed5b4","prompt_sha256":"e4f56bf568e7854d0a1d2efb6ec96f8697d5132737cdb31cb7fc64acbd923974","message_sha256":"4cbe95ed20956d02a1cdd4f467f523197758b0b579b0f7f7ccb088d602a88544","cache_read_input_tokens":0,"cache_creation_input_tokens":0}},"version":"pith-run-record/v1","prompt_set_sha256":"d60f54226f003b7308ab6f0f9f2951c09c726f00cc66fc05af3456f212247e93","configured_pipeline_version":"pith-pipeline@v1.4.1-daily-deepseek-or-medium"},"basis":{"kind":"full_text","source":"recorded","chars_given":88019,"chars_total":88019},"receipt":{"review_sha256":"0bffabe5b4261f9d3df8a10a7be5ad68f6113cd101afe0ed244ca6487c48e2d6","rule":"pith-review-sha256/v1","hashed_at":"2026-11-20T06:27:58.013480+00:00","fields":["id","arxiv_id","paper_version","created_at","pipeline_version","model_set","run_record","input_kind","input_chars","input_total_chars","passes_missing","verdict","confidence","novelty_score","correctness_risk","formal_verification","parameter_count","discovery_kind","one_line_summary","lead","skeptic_critique","referee_report","author_rebuttal","desk_editor","pith_extraction","objection_audit","published_standing","published_confidence"],"matches_current":true},"falsifier":"Run AutoMaster with the same teacher, task pool, and SFT recipe but with the agent's skill design and data selection replaced by random selection of 512 trajectories from the same pool of unguided successful rollouts, and compare to Vanilla SFT on a single fixed run. If the gains disappear, the headroom is due to selection rather than behavioral shaping.","supporting_citations":[],"lane_accuracy":{"status":"not yet measured","lane":{"model":"deepseek/deepseek-v4-flash","input_kind":"tex"},"url":"https://pith.science/accuracy"},"review_version":1}