Machine Learning · arXiv:2609.00196 v1 · submitted 31 Aug 2026 · 29 references
WHALE: A Simple Recipe for Joint Harness-Weight Optimization
The paper says
“Agent performance depends jointly on the model parameters and the executable harness code that manages context and control flow.”
more of the abstract
Optimizing either component in isolation can leave the system bottlenecked by its frozen counterpart: weight updates can change which harness is effective, while harness updates can change which model capabilities are exposed. Existing joint-adaptation methods optimize weights and textual prompts but leave the broader harness fixed. We propose Weight-Harness Alternating LEarning (WHALE), a simple recipe that alternates two phases: updating the model under the current harness, then searching for a better harness under the updated model. We instantiate these two phases with online rejection-sampling fine-tuning and Meta-Harness, respectively. When to switch is a key design choice: to separate real improvements from noise without over-optimizing against a changing counterpart, WHALE uses either fixed phase durations or an adaptive patience rule over training signals. Using Qwen3.5-2B/4B agents across three domains (search question answering, mathematical reasoning, and chess puzzles), WHALE outperforms weight-only, harness-only, and Fast-Slow Training by 4.15-24.38 percentage points in best mean@8 accuracy. Either component can be the bottleneck: harness search matches peak weight-only accuracy with far fewer rollouts in SearchQA, but improves math accuracy only after a weight update. Small interleaved updates also outperform stagewise weight-then-harness optimization in accuracy and rollout cost. The code is available at https://github.com/krafton-ai/WHALE.
from the abstract
Pith says
Machine review · deepseek-v4-flash · 9 Sep 2026 · what was read was not recorded · not checked by a person
It can be wrong. Challenge this review or reply to it.
Holds up, with conditions. A clean alternating optimization recipe let down by a test-set selection protocol that makes the headline numbers uninterpretable as reported. This line is wrong
4 major objections · 6 minor
How often objections from reviews like this one are right: not yet measured
Not checked: figures, code, what the 29 references say.
The paper says, at Algorithm 1 and §5.2, Fig. 2, Table 3
The review did not quote this passage.
Pith says: Objection 1 of 4 · major
The evaluation protocol uses the same test set for both checkpoint selection and final reporting. Algorithm 1 (line 17) returns the pair with the best accuracy on D_test, and Fig. 2 and Table 3 report 'best test mean@8' over the run.
more of objection 1
Since mean@8 is a noisy estimate (e.g., only 60 test items for Math), taking the maximum over 6–10 cycles introduces optimistic selection bias that is not necessarily equal across methods, because methods have different numbers of candidate checkpoints and different trajectory shapes. The central quantitative claim—that WHALE beats baselines by 4.15–24.38 pp—is therefore not yet established as a property of the alternating recipe; part of the reported gain may be a selection artifact. The authors should either (a) use a held-out validation set for checkpoint selection and report accuracy on a separate test set, (b) report final-checkpoint accuracy instead of best-over-run, or (c) characterize the bias magnitude through, e.g., a control experiment on a fixed checkpoint schedule.
The paper says, at §5.3 (FST comparison)
The review did not quote this passage.
Pith says: Objection 2 of 4 · major
The FST comparison replaces the original RL weight update with RSFT and restricts Meta-Harness to prompt search. While the authors argue this isolates the effect of the harness search space, the resulting system is no longer the published Fast-Slow Training method.
more of objection 2
This is not a fatal flaw—it is a reasonable controlled comparison—but the paper should be more explicit in the main text (not just Appendix B.1) that this is a modified variant, not the original FST, and discuss whether the original FST's RL update could benefit from the joint prompt-weight setting in ways that RSFT cannot capture.
The paper says, at §6.1 (bottleneck analysis)
The review did not quote this passage.
Pith says: Objection 3 of 4 · major
The bottleneck analysis in §6.1 relies on the same test-set trajectories used for checkpoint selection. If the selected checkpoint systematically overfits to test-set idiosyncrasies (e.g., format compliance patterns in SearchQA), the behavioral metric analysis may reflect spurious correlations rather than genuine regime properties. The authors should confirm that the qualitative findings (harness-dominant vs. model-dominant, catalytic effect of small weight updates) hold when evaluated with a proper train/validation/test split.
The paper says, at §6.2.3 and Fig. 6 (adaptive WHALE)
The review did not quote this passage.
Pith says: Objection 4 of 4 · major
The adaptive WHALE results in SearchQA (52.82%) and Math (26.46%) are compared against fixed-schedule baselines that themselves were selected based on test-set accuracy (best-over-run). Gien the test-set selection bias, it is unclear whether adaptive WHALE's superiority is genuine or an artifact of different selection dynamics. Moreover, the adaptive rule uses training signals (valid) but the final checkpoint is still chosen on the test set. The comparison should be repeated with a validation-based selection protocol.
6 minor objections
The paper says, at Algorithm 1, line 17
Pith says: minor objection 1 of 6
The return statement selects on D_test. This is stated transparently, but the paper should explicitly discuss the bias this introduces and why it is or is not acceptable (currently no such discussion exists).
The paper says, at Appendix B.1 (Table 1)
Pith says: minor objection 2 of 6
The table reports 'Completed cycles' for WHALE: SearchQA 6, Math 9, Chess 6. The number of cycles differs across domains, but the authors do not explain why Math uses 9 cycles while the others use 6. If this was determined by a fixed budget or by a stopping rule, that should be stated.
show 4 more
The paper says, at Fig. 2
Pith says: minor objection 3 of 6
The 'best-so-far' curves in Fig. 2 are visually useful but would benefit from error bands (e.g., bootstrap confidence intervals) to convey the uncertainty in the mean@8 estimates, especially for Math where there are only 60 test examples.
The paper says, at §1, Abstract
Pith says: minor objection 4 of 6
Abstract and intro claim gains of '4.15–24.38 percentage points' but the range spans a factor of 6. This wide range suggests the benefit is heterogeneous across domains; the authors should discuss what drives the extremes (Chess apparently gives the lower bound, SearchQA the higher).
The paper says, at §4, Algorithm 2
Pith says: minor objection 5 of 6
Algorithm 2 introduces per-phase patience parameters (W, T_min, P_w, J_min, P_h) that are set by default to (0.2 epoch, 0.2 epoch, 0.2 epoch, 6, 2). These are hyperparameters, so the adaptive WHALE still requires tuning, contrary to the claim that it 'removes the schedule (E,I) from the hyperparameters.' The authors should clarify that it removes the per-cycle fixed schedule but introduces new patience hyperparameters.
The paper says, at §5.1 (Ches Puzzles)
Pith says: minor objection 6 of 6
The Chess Puzzles domain uses Qwen3.5-4B (vs. 2B for the other two domains). The authors should discuss whether model capacity interacts with the benefits of joint optimization, since the observation that Chess is 'model-dominant' might be partially an artifact of using a larger model.
What would settle it
Hold out a selection split disjoint from the headline test set (e.g., 30% of D_test items per domain) and forbid its use in reported numbers. For each method and seed, choose the final (θ,h) by best mean@8 on the selection split (or by training reward for adaptive WHALE), then evaluate that single chosen pair once on the untouched 70% and average over 3 seeds. If WHALE's margin over the stronger single-component baseline remains ≥7.67 pp and over the prompt-restricted control ≥4.15 pp, the concern is resolved; if the margin shrinks materially or flips, the central claim should be revised to validation-selected accuracy.
from the skeptic's reading
How the review went
4 readings by deepseek-v4-flash · 9 Sep 2026
- Reader holds up, with conditions
- Skeptic disagrees
- Editor wrote the verdict above
- Referee major revision: 4 major objections, 6 minor
Not run on this paper: the simulated authors' defence.
Pith did not check this review's objections against the paper.
The review in full
The review in brief, machine-written
The core of WHALE is a natural idea: alternate RSFT weight updates with Meta-Harness code search, treating the harness as an optimizable artifact alongside the weights. The paper executes this idea cleanly — matched budgets, per-domain breakdowns, bottleneck analysis that actually separates harness-dominant from weight-dominant regimes, and released code. The schedule sensitivity analysis and the adaptive patience rule are thoughtful. If the central claim holds, this is a useful recipe for the agent-optimization toolkit.
Here is the honest bottom line: I cannot trust the quantitative headline. Algorithm 1 selects the checkpoint that performed best on D_test, and Figure 2 and Table 3 report "best test mean@8" over the run. Mean@8 is a noisy estimate (8 rollouts per example; Math has only 60 test items), and taking the maximum over 6–10 cycles guarantees an upward bias. The bias is not constant across methods — different numbers of candidate checkpoints and different curve shapes mean different selection pressures. So the claim that WHALE beats weight-only, harness-only, and FST by 4.15–24.38 percentage points is not yet established as a property of the alternating recipe. To their credit, the paper is transparent about the protocol, but that does not make the bias go away.
The reader flags the modified FST comparison and the lack of error bars. Both are real but secondary. The FST modification is a restricted ablation, not a direct comparison to the published method, and the abstract's "outperforms FST" is misleading as written. But even a faithful FST baseline would not fix the test-set selection problem.
The paper otherwise shows solid engineering and careful analysis. The bottleneck regimes and the schedule study are informative even without the absolute accuracy numbers. This paper deserves a serious referee — the core idea is real — but the authors need to re-run their analysis with a held-out validation split or bootstrap-corrected selection before the quantitative claims can be taken at face value.
The strongest objection
The most load-bearing issue is the evaluation protocol, not search expressivity. Algorithm 1 (line 17) returns '(θ_k,h_k) with the best accuracy on D_test,' and Fig. 2 and Table 3 report 'best test mean@8' over the run. No validation split is used for checkpoint or cycle selection, so the same D_test both selects and evaluates. Since mean@8 is a noisy estimate (8 rollouts per example; Math has only 60 test items), taking the maximum over 6–10 cycles biases the reported accuracy upward, and the bias is not necessarily equal across methods because WHALE and the baselines have different numbers of candidate checkpoints and different curve shapes. Thus the central quantitative claim—WHALE beats weight-only, harness-only, and prompt-restricted FST by 4.15–24.38 pp—is not yet established as a property of the alternating recipe; part of the reported gain may be a selection artifact. The reader's FST and seed concerns are real but secondary; even a faithful FST baseline would not fix test-set selection. The issue is transparently stated in the algorithm, so it is directly checkable.
The reader's verdict is CONDITIONAL, and I keep it conditional, but for a different reason. The strongest claim rests on 'best mean@8' numbers, and Algorithm 1 makes explicit that this best is selected on D_test. That means the reported accuracy is not the accuracy of a model chosen by a reproducible selection rule; it is the maximum of several noisy estimates evaluated on the very data used for the claim. This is most dangerous in Math (60 test items) and Chess (256 items), and it can affect comparisons because the methods have different numbers of candidate checkpoints. I do not consider this a sign of bad faith; the protocol is stated openly, which makes the issue checkable. But it is a correctness risk in the central claim. The proposed check is cheap and decisive: use a validation split for selection and keep D_test untouched. If the margins survive, the concern is resolved; if not, the paper needs new experiments. The reader's weakest assumption about harness search space and proposer quality is legitimate but less load-bearing: a narrower search space would weaken generality, whereas test-set selection threatens the specific numbers in the abstract. The FST implementation caveat and lack of seeds remain worth fixing; a held-out selection run with 3 seeds addresses the seed issue and the selection issue simultaneously, and the FST label should be corrected to 'prompt-restricted control' in the abstract as the reader suggested.
The referee's report
The problem of jointly optimizing model parameters and executable harness code is timely and practically important for agentic LLM systems. The observation that the bottleneck can shift between model and harness, and that small weight updates can 'unlock' otherwise ineffective harness search, is potentially valuable guidance for practitioners. If the central claims held, the paper would provide a clear, modular recipe for a significant improvement over existing methods. However, as detailed below, a methodological flaw in the evaluation protocol undermines confidence in the reported margins.
Pith's reading of the paper (a machine's summary, not the authors')
An agentic language model's usefulness is set by two coupled components: the model weights, and the harness, the executable code that decides what enters context, how tools are called, how errors are handled, and when to stop. Optimizing one while freezing the other, the paper argues, leaves the system bottlenecked by its frozen half, and prior joint methods only adapted textual prompts, not the broader harness. WHALE instead alternates small rejection-sampling fine-tuning of the weights with Meta-Harness search over full harness programs, and reports gains of 4.15 to 24.38 percentage points in best mean@8 accuracy over weight-only, harness-only, and prompt-weight joint optimization across search QA, math reasoning, and chess puzzles. The paper's deeper claim is that a model and its harness should be treated as one jointly trained system, with the timing of switches between the two phases as the decisive design choice.
What carries the argument. The load-bearing mechanism is alternation applied to the coupled objective $J(\theta,h)$ over model parameters $\theta$ and executable harness programs $h$, where the two are linked only through the trajectory distribution $\pi_{\theta,h}$ they induce. Each cycle holds one side fixed while updating the other, $\theta_{k+1} = \text{ModelUpdate}(\theta_k; h_k)$ and then $h_{k+1} = \text{HarnessSearch}(h_k; \theta_{k+1})$, a coordinate-descent structure that keeps a changing counterpart from confounding credit assignment. The weight phase is online rejection-sampling fine-tuning, supervised learning on the model's own verifier-accepted rollouts; the harness phase is Meta-Harness, an iterative propose–evaluate–select search over harness code that keeps an archive of candidates so later proposals can refine or revert. The third ingredient is the switching schedule: too little evidence per phase adopts a noisy candidate and can destabilize training, while too much over-optimizes against a counterpart that will change, and a per-phase patience rule over training signals replaces the fixed budgets.
Core claim. WHALE's central claim is that the joint objective $J(\theta,h)$ — expected verifier reward over trajectories produced by the pair $(\theta,h)$ — is best optimized not by one long pass per component but by short alternating conditional updates, and that the harness side of the pair should be the full executable harness, not just the prompt text. With weights updated by online rejection-sampling fine-tuning and harnesses searched by Meta-Harness, one fixed schedule $(E,I)=(0.6,6)$ beats weight-only, harness-only, and a prompt-restricted Fast–Slow Training control in all three domains, improving on the stronger single-component baseline by 7.67–10.05 percentage points and on Fast–Slow Training by 4.15–13.00 points. The experiments identify two bottleneck regimes: SearchQA is harness-dominant, where harness search matches peak weight-only accuracy with 5.79% of the rollouts, while Math is model-dominant, where harness search alone scores almost nothing until a small weight update makes it effective. Stagewise weight-then-harness optimization is dominated by small alternating steps (5.32 and 9.16 percentage points behind in the two domains), and an adaptive patience rule that stops each phase when its training signal stops improving reaches the best overall result in SearchQA and trails only the best hand-tuned schedule in Math.
The premise it rests on. The result rests on the search space for harnesses and the proposer that generates harness candidates being strong enough to keep finding genuinely better programs than the hand-written starting harness, and the paper does not measure how WHALE degrades as either is weakened.
If the paper is right (Pith's inferences, not the authors')
In harness-dominant domains such as SearchQA, a short harness search matches or beats extensive weight training at a small fraction of the rollout cost, which makes harness search a cheap diagnostic to run before committing to model training.
In model-dominant domains such as Math, the weight phase is a prerequisite that unlocks harness search, so the order and interleaving of phases determines whether the harness side contributes anything at all.
One long weight pass followed by one long harness pass is dominated by small alternating updates, both in final test accuracy and in the rollouts needed to reach a target accuracy.
The same interleaving schedule, with small weight epochs and harness search long enough to beat noise, is near-optimal across domains at opposite bottleneck extremes, and the adaptive patience rule recovers that schedule without hand-tuned phase budgets.
Because WHALE exposes only black-box interfaces for the two phases, the recipe transfers to any weight-update and harness-search algorithm that share a fixed binary verifier.
A robustness question the paper leaves open is how much of the gain belongs to alternation versus to the proposer used inside Meta-Harness; replacing the proposer with a much weaker model at matched budget would separate the two, since the paper does not characterize how performance degrades along that axis.
The per-domain initial harnesses are hand-designed, so the recipe presumes a reasonable starting point; starting WHALE from a deliberately bad or randomized harness would test whether alternation climbs on its own or mostly polishes a good head start.
The SearchQA trajectories show the searched harness writing turn-by-turn state lines, answer checklists, and stricter termination rules, which suggests harness search is partly automating prompt engineering inside the loop; restricting the search space to exactly those prompt-like fields would quantify how much of WHALE's edge survives.
The token-budget logic of the Math and Chess examples suggests the joint objective also implicitly rewards cost savings, not just accuracy; measuring tokens-per-solved-example under WHALE versus weight-only training would make that trade-off explicit.
Check this review
Pith stored no record of this review's passes and took no fingerprint of it. Reviews made by code from 13 Nov 2026 on carry both.
How often reviews like this one are right
Reviews like this one: written by deepseek-v4-flash, where what was read was not recorded.
How often objections from reviews like this one are right: not yet measured
Reviews by people
This review was written by a machine. No person has reviewed this paper on Pith yet. A scientist listed on Pith can review it for a fee, and their review is published here whatever it concludes.
The paper
Abstract
Agent performance depends jointly on the model parameters and the executable harness code that manages context and control flow. Optimizing either component in isolation can leave the system bottlenecked by its frozen counterpart: weight updates can change which harness is effective, while harness updates can change which model capabilities are exposed. Existing joint-adaptation methods optimize weights and textual prompts but leave the broader harness fixed. We propose Weight-Harness Alternating LEarning (WHALE), a simple recipe that alternates two phases: updating the model under the current harness, then searching for a better harness under the updated model. We instantiate these two phases with online rejection-sampling fine-tuning and Meta-Harness, respectively. When to switch is a key design choice: to separate real improvements from noise without over-optimizing against a changing counterpart, WHALE uses either fixed phase durations or an adaptive patience rule over training signals. Using Qwen3.5-2B/4B agents across three domains (search question answering, mathematical reasoning, and chess puzzles), WHALE outperforms weight-only, harness-only, and Fast-Slow Training by 4.15-24.38 percentage points in best mean@8 accuracy. Either component can be the bottleneck: harness search matches peak weight-only accuracy with far fewer rollouts in SearchQA, but improves math accuracy only after a weight update. Small interleaved updates also outperform stagewise weight-then-harness optimization in accuracy and rollout cost. The code is available at https://github.com/krafton-ai/WHALE.
The paper's own sentences, arranged, kept beside the review and separate from it. Each label is drawn from a closed vocabulary; the number beside it is the probability the classifier gave that label. No sentence in this record was written by a model.
abstract compiled · 2026-09-23
The authors' main claim, as the classifier picked it out 1.00
“Using Qwen3.5-2B/4B agents across three domains (search question answering, mathematical reasoning, and chess puzzles), WHALE outperforms weight-only, harness-only, and Fast-Slow Training by 4.15-24.38 percentage points in best mean@8 accuracy.”
How the main claim is stated: asserted 1.00
What the abstract itself states, as probabilities: claims first 0.12 · states limitation 0.36 · human participants 0.03 · controlled comparison 0.95 · randomized assignment 0.02 · states quantitative result 0.99 · compares with prior results 0.98 · mentions code or data availability 0.98
The abstract, sentence by sentence
This table labels the abstract only (arXiv listing abstract v1). The number is the probability the classifier gave the label.
| Sentence | Role | The authors' own work | States the main conclusion | Strength |
|---|---|---|---|---|
| Agent performance depends jointly on the model parameters and the executable harness code that manages context and control flow. | background 1.00 | 0.12 | 0.05 | asserted 1.00 |
| Optimizing either component in isolation can leave the system bottlenecked by its frozen counterpart: weight updates can change which harness is effective, while harness updates can change which model capabilities are exposed. | background 0.79 | 0.15 | 0.09 | asserted 0.55 |
| Existing joint-adaptation methods optimize weights and textual prompts but leave the broader harness fixed. | gap 0.96 | 0.05 | 0.04 | asserted 1.00 |
| We propose Weight-Harness Alternating LEarning (WHALE), a simple recipe that alternates two phases: updating the model under the current harness, then searching for a better harness under the updated model. | method 0.51 | 0.84 | 0.19 | asserted 1.00 |
| We instantiate these two phases with online rejection-sampling fine-tuning and Meta-Harness, respectively. | method 0.97 | 0.85 | 0.04 | asserted 0.99 |
| When to switch is a key design choice: to separate real improvements from noise without over-optimizing against a changing counterpart, WHALE uses either fixed phase durations or an adaptive patience rule over training signals. | method 0.98 | 0.79 | 0.05 | asserted 0.99 |
| Using Qwen3.5-2B/4B agents across three domains (search question answering, mathematical reasoning, and chess puzzles), WHALE outperforms weight-only, harness-only, and Fast-Slow Training by 4.15-24.38 percentage points in best mean@8 accuracy. | result 1.00 | 0.97 | 0.73 | asserted 1.00 |
| Either component can be the bottleneck: harness search matches peak weight-only accuracy with far fewer rollouts in SearchQA, but improves math accuracy only after a weight update. | result 0.70 | 0.94 | 0.08 | asserted 0.98 |
| Small interleaved updates also outperform stagewise weight-then-harness optimization in accuracy and rollout cost. | result 1.00 | 0.93 | 0.10 | asserted 1.00 |
| The code is available at https://github.com/krafton-ai/WHALE. | contribution 0.77 | 0.34 | 0.02 | asserted 0.90 |
What kind of paper the classifier reports this is: method or tool 1.00 · evidence from observation or experiment 0.95
What was examined, and what was not
Sources read
- arXiv listing abstract v1
Not examined
- full text
- figures and tables
- references and citing sentences
- relations to other papers
Levels run
- P abstract
Not run
- S sections
- T body sentences
- C citations
- R cross-paper relations
How the labels were made
Compiled by classifier-1.13.0 via TypeSafe API, question bank abstract-p1-20260923, 2026-09-23: 1 request, 7,660 input tokens.
Every quoted sentence on this record is the authors'. Every label is from Pith's closed vocabulary, with the probability shown beside it. This record as JSON
Figures (7)
Figures are the authors', reproduced from the paper.
References (29)
- [1] B.3.1 SEARCHQA LLM-AS-A-JUDGEPROMPT The SearchQA judge prompt in full. Placeholders are filled per call with the question, the reference answers, and the extracted model answer. You are an impartial judge for a question-answering task. Mark CORRECT when predicted answer refers... on Pith
- [3] Jiazhan Feng, Shijue Huang, Xingwei Qu, Ge Zhang, Yujia Qin, Baoquan Zhong, Chengquan Jiang, Jinxin Chi, and Wanjun Zhong. Retool: Reinforcement learning for strategic tool use in llms. arXiv preprint arXiv:2504.11536, on PithPith's reviewarXiv
- [4] Table 3: Best test mean@8 accuracy (%) of each run in Figure 4 and the rollouts consumed up to that point. Bold denotes the best accuracy in each domain and underlining the second best. SearchQA Math Runmean@8Rolloutsmean@8Rollouts Weight-only 38.27 530,432 15.42 645,120 Harne... on PitharXiv
- [6] URL https: //arxiv.org/abs/2408.08435. Bowen Jin, Hansi Zeng, Zhenrui Yue, Dong Wang, Hamed Zamani, and Jiawei Han. Search-r1: Training llms to reason and leverage search engines with reinforcement learning.arXiv preprint arXiv:2503.09516, on PithPith's reviewarXiv
- [7] Jeff Johnson, Matthijs Douze, and Herv ´e J´egou. Billion-scale similarity search with gpus.arXiv preprint arXiv:1702.08734, on PithPith's reviewarXiv
- [9] Ac- cessed 2026-07-20. Jiahang Lin, Shichun Liu, Chengjun Pan, Lizhi Lin, Shihan Dou, Zhiheng Xi, Xuanjing Huang, Hang Yan, Zhenhua Han, Tao Gui, et al. Agentic harness engineering: Observability-driven automatic evolution of coding-agent harnesses.arXiv preprint arXiv:2604.25850, on PithPith's reviewarXiv
- [10] Xinyu Lu, Kaiqi Zhang, Jinglin Yang, Boxi Cao, Yaojie Lu, Hongyu Lin, Min He, Xianpei Han, and Le Sun. Pˆ 2o: Joint policy and prompt optimization.arXiv preprint arXiv:2603.21877, on PithPith's reviewarXiv
- [13] URLhttps://arxiv.org/abs/2606.05922. Sungho Park, Wonjoong Kim, Rongyuan Tan, Jue Zhang, Wook-Shin Han, Pengfei Gao, Chanyoung Park, Yongqiang Yao, Rao Fu, Elsie Nallipogu, Qingwei Lin, Saravan Rajmohan, and Dongmei Zhang. Autosaddler: Automatic harness optimization with durab... on PithPith's reviewarXiv
- [14] URLhttps://arxiv.org/abs/2608.23041. Ofir Press, Muru Zhang, Sewon Min, Ludwig Schmidt, Noah A. Smith, and Mike Lewis. Measuring and narrowing the compositionality gap in language models.Findings of the Association for Computational Linguistics: EMNLP, on PithPith's reviewarXiv
- [15] URL https: //arxiv.org/abs/2604.20938. Yu Shang, Yu Li, Keyu Zhao, Likai Ma, Jiahe Liu, Fengli Xu, and Yong Li. Agentsquare: Automatic llm agent search in modular design space, on PithPith's reviewarXiv
- [16] URL https://arxiv.org/abs/2410. 06153. Zhihong Shao, Peiyi Wang, Qihao Zhu, Runxin Xu, Junxiao Song, Xiao Bi, Haowei Zhang, Mingchuan Zhang, Y . K. Li, Y . Wu, and Daya Guo. Deepseekmath: Pushing the limits of mathematical reasoning in open language models.arXiv preprint arXiv... on PithPith's reviewarXiv
- [17] URL https: //arxiv.org/abs/2312.06585. Dilara Soylu, Christopher Potts, and Omar Khattab. Fine-tuning and prompt optimization: Two great steps that work better together.arXiv preprint arXiv:2407.10930, on PithPith's reviewarXiv
- [18] 12 Rishabh Tiwari, Kusha Sareen, Lakshya A. Agrawal, Joseph E. Gonzalez, Matei Zaharia, Kurt Keutzer, Inderjit S. Dhillon, Rishabh Agarwal, and Devvrit Khatri. Learning, fast and slow: Towards LLMs that adapt continually.arXiv preprint arXiv:2605.12484, on PithPith's reviewarXiv
- [20] Accessed 2026-07-20. Qiying Yu, Zheng Zhang, Ruofei Zhu, Yufeng Yuan, Xiaochen Zuo, Yu Yue, Tiantian Fan, Gaohong Liu, Lingjun Liu, Xin Liu, Haibin Lin, Zhiqi Lin, Bole Ma, Guangming Sheng, Yuxuan Tong, Chi Zhang, Mofan Zhang, Wang Zhang, Hang Zhu, Jinhua Zhu, Jiaze Chen, Jian... on PithPith's reviewarXiv
- [21] doi: 10.52202/ 075280-0482. URL https://proceedings.neurips.cc/paper_files/paper/ 2023/file/23e6f78bdec844a9f7b6c957de2aae91-Paper-Conference.pdf. Eric Zelikman, Yuhuai Wu, Jesse Mu, and Noah D. Goodman. Star: Bootstrapping reasoning with reasoning, on Pith
- [22] URLhttps://arxiv.org/abs/2203.14465. Jiayi Zhang, Jinyu Xiang, Zhaoyang Yu, Fengwei Teng, Xionghui Chen, Jiaqi Chen, Mingchen Zhuge, Xin Cheng, Sirui Hong, Jinlin Wang, Bingnan Zheng, Bang Liu, Yuyu Luo, and Chenglin Wu. Aflow: Automating agentic workflow generation.arXiv prep... on PithPith's reviewarXiv
- [23] Lunjun Zhang, Ryan Chen, and Bradly C. Stadie. Evolutionary system prompt learning for reinforce- ment learning in LLMs.arXiv preprint arXiv:2602.14697, on PithPith's reviewarXiv
- [24] Mingchen Zhuge, Wenyi Wang, Louis Kirsch, Francesco Faccio, Dmitrii Khizbullin, and J ¨urgen Schmidhuber. Language agents as optimizable graphs.arXiv preprint arXiv:2402.16823, on PithPith's reviewarXiv
- [25] SearchQA.The harness search may additionally rewrite or expand a model-generated query before retrieval and set retrieval parameters and the ranking of retrieved documents. Each tool call accepts one query, and at most 16 assistant turns are allowed. The initial harness instru... on Pith
- [28] Then subtract 2025 from this number.,→ Let the digits be $x_1, x_2, x_3, x_4, x_5, x_6, x_7, x_8$ in some order. The digits are distinct and form the sequence $1, 2, 3, 4, 5, 6, 7, 8$. The number must be divisible by on Pith
show 9 more
- [279] So answer = 2304 - 2025 = on Pith
- [2015] Chengrun Yang, Xuezhi Wang, Yifeng Lu, Hanxiao Liu, Quoc V . Le, Denny Zhou, and Xinyun Chen. Large language models as optimizers.arXiv preprint arXiv:2309.03409, 2024a. John Yang, Carlos Jimenez, Alexander Wettig, Kilian Lieret, Shunyu Yao, Karthik Narasimhan, and Ofir Press.... on PithPith's reviewarXiv
- [2019] Yoonho Lee, Roshen Nair, Qizheng Zhang, Kangwook Lee, Omar Khattab, and Chelsea Finn. Meta-harness: End-to-end optimization of model harnesses.arXiv preprint arXiv:2603.28052, on PithPith's reviewarXiv
- [2022] URL https://arxiv. org/abs/2112.09332. Wenbo Pan, Shujie Liu, Chin-Yew Lin, Jingying Zeng, Xianfeng Tang, Xiangyang Zhou, Yan Lu, and Xiaohua Jia. Evolving agents in the dark: Retrospective harness optimization via self-preference, on PithPith's reviewarXiv
- [2023] URLhttps://arxiv.org/abs/2310.05915. Hanze Dong, Wei Xiong, Deepanshu Goyal, Yihan Zhang, Winnie Chow, Rui Pan, Shizhe Diao, Jipeng Zhang, KaShun Shum, and Tong Zhang. RAFT: Reward ranked finetuning for generative foundation model alignment.Transactions on Machine Learning Research, on PithPith's reviewarXiv
- [2024] Accessed 2026-07-20. Reiichiro Nakano, Jacob Hilton, Suchir Balaji, Jeff Wu, Long Ouyang, Christina Kim, Christopher Hesse, Shantanu Jain, Vineet Kosaraju, William Saunders, Xu Jiang, Karl Cobbe, Tyna Eloundou, Gretchen Krueger, Kevin Button, Matthew Knight, Benjamin Chess, an... on Pith
- [2025] Caglar Gulcehre, Tom Le Paine, Srivatsan Srinivasan, Ksenia Konyushkova, Lotte Weerts, Abhishek Sharma, Aditya Siddhant, Alex Ahern, Miaosen Wang, Chenjie Gu, Wolfgang Macherey, Arnaud Doucet, Orhan Firat, and Nando de Freitas. Reinforced self-training (ReST) for language mode... on PithPith's reviewarXiv
- [2026] Xiaohe Bo, Rui Li, Zexu Sun, Quanyu Dai, Zeyu Zhang, Zihang Tian, Xu Chen, and Zhenhua Dong. Prompt and parameter co-optimization for large language models. InInternational Conference on Learning Representations, volume 2026, pp. 116697–116723, on Pith
- [2304] N - 2025 = 27 [... trimmed ...] A number is divisible by 22 if it's divisible by 2 and on Pith
Cited by 14 Pith papers
Reviewed papers on Pith that cite this one. OpenAlex reports about 17 citations worldwide. The full citation record
- HandRSI: Recursive Self-Improvement for Hand Pose Estimation cs.CV · Nov 2026
- ModHarness-RL: Recasting Harnesses as Learnable and Evolving Programs for Agentic RL cs.AI · Nov 2026
- JoMoHa: Joint Model--Harness Optimization with Reinforcement Learning cs.LG · Nov 2026
- CoEvo-RSI: When Harnesses Evolve, Which Models Should Follow? cs.LG · Oct 2026
- Harness Evolution Hits a Ceiling: When Weight Training Should Begin cs.AI · Oct 2026
- Not All Experience Belongs in the Weights: Component Routing for Self-Improving GUI Agents cs.AI · Oct 2026
- Where Should LLM Agents Adapt at Test Time? In Context, in Weights, or Both cs.LG · Nov 2026
- Where Should Learning Live? Regret Analysis for Text, Parameter, and Joint Adaptation stat.ML · Nov 2026
- CoHARL: Co-Evolving Agent Policies and Executable Harnesses with Reinforcement Learning cs.LG · Oct 2026
- Elicitation or Expansion? Demystifying Gains from Agent Harness Optimization cs.LG · Oct 2026
- ActiveSaddler: Automated Curriculum Learning for Agent Harness Optimization cs.AI · Oct 2026
- AgentVidBench: A Multi-Hop Video Question Answering Benchmark for Evaluating MLLM Agents cs.CV · Sep 2026
- VACE: Validation-Gated Alternating Co-Evolution of Agent Models and Harnesses cs.LG · Sep 2026
- PUBG Ally: A Conversational Embodied Agent as an AI Teammate cs.AI · Sep 2026
Challenges and discussion
Disagree with the review, or know something it missed? Leave a comment, or challenge this review formally: a challenge becomes a permanent entry in this paper's record, and its decision is published here.
Come back by email
A short list of the week's newly reviewed papers. Nothing is sent until you confirm from your inbox, and every email has a link that stops it. No account.
No comments yet.
Sign in to comment