Pith. sign in

REVIEW 4 major objections 4 minor 49 references

Post-Training on Office Work Improves Software Engineering: A Behavioral Account of Cross-Domain Transfer

T0 review · 4 major / 4 minor · reviewed 2026-08-15 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read Office-workflow training lifts coding benchmark by 5.8 points

desk verdict The +5.8pp transfer is plausible but rests on a single unseeded run; the GDE framing adds interpretation rather than evidence, and the paper needs multi-seed replication and a reconciliation of its reported counts before the result is trustworthy. read the letter →

arxiv 2608.01604 v1 pith:OHPSZPOP submitted 2026-08-03 cs.AI cs.SE

Logan Ritchie , Sushant Mehta , Liudas Panavas , Edwin Chen This is my paper · ORCID
classification cs.AIcs.SE
keywords goal-directedexecutioncross-domaintransferlong-horizonpost-trainingSWE-BenchProofficeworkflowsreinforcementlearningagenttrajectoriesbehavioralanalysis
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

The paper tries to establish that long-horizon post-training on tasks from one domain can strengthen a domain-general capability, which it calls goal-directed execution (GDE), rather than only carving task-specific skills. It post-trains a 122-billion-parameter model on 363 office-workflow tasks that contain no software-engineering content, then measures the model on the SWE-Bench Pro coding benchmark: first-attempt success rises from 20.5% to 26.3%. Paired trajectory analysis of tasks that the base model failed and the trained model passed shows gains in all four GDE behaviors—goal formation, state construction, goal stability, and verification—in both office and software settings. The causal link to long-horizon task structure is presented as a hypothesis, but if the empirical transfer holds, it implies that data-curation and post-training decisions should account for the structural demands a task exercises, not just its subject matter.

What carries the argument

The carrying mechanism is a recursive goal-loop model of agent behavior, formalized as goal-directed execution (GDE): the agent forms a goal from its parent objective and working state, acts, updates its working state from environment feedback, and verifies whether the state satisfies the goal, decomposing into nested loops when actions are too abstract. The four capabilities named by GDE—goal formation, state construction, goal stability, and verification—are the interpretive grid used to compare base and trained trajectories. Training itself runs through a two-stage post-training recipe on 363 LHMTA office tasks: a supervised warm-up on high-scoring teacher trajectories, followed by GSPO reinforcement learning with a dense reward equal to the fraction of grader criteria satisfied per trajectory. The design isolates the structural demand of long-horizon work because the training tasks share no content with software engineering.

What would settle it

An audit would settle it: if the 363-task LHMTA snapshot or the supervised teacher trajectories contain repository content, tests, or grader outputs from SWE-Bench Pro tasks, or if retraining with the same data but selecting the checkpoint on the LHMTA holdout alone fails to reproduce the roughly 5.8-point SWE-Bench Pro gain, the transfer claim as stated would not survive. A cleaner experiment is to repeat the full two-stage run with a held-out copy of SWE-Bench Pro never opened during development and pre-register the checkpoint-selection rule.

Watch

Extended reading notes

Core claim

The central discovery is that a model trained entirely on office workflows—documents, spreadsheets, web research, file manipulation, and scheduling—improves at resolving software-repository issues. The trained checkpoint scores 26.3% pass@1 on SWE-Bench Pro under greedy decoding versus 20.5% for the base model, a gain of 5.8 percentage points, despite the training collection containing no software-engineering tasks, graders, or benchmark feedback. The paper argues that the gain is best explained behaviorally: post-training strengthens four observable capabilities that transfer across domains—forming the right next goal, constructing and maintaining task-relevant state, preserving higher-level requirements under local pressure, and verifying completion against the environment. Matched trajectory analysis of 103 newly-passing tasks shows these differences in both domains, and aggregate SWE-Bench Pro metrics shift in the same direction: less repeated retrieval, more contact with reference-patch files, far smaller patches, and nearly double the share of runs that execute a formal test.

Load-bearing premise

The load-bearing premise is that the SWE-Bench Pro coding benchmark never influenced the training run: no benchmark tasks or grader outputs entered the 363-task collection, and no checkpoint, hyperparameter, or reward choice was made using benchmark scores.

Editorial extensions

If this is right

  • Post-training data value should be measured by the behavioral demands it exercises, not only by topic: a domain with no coding content can improve a coding benchmark.
  • The two-stage recipe—supervised warm-up plus GSPO with dense criterion-level rewards—produces transfer beyond the training environment: LHMTA holdout +17.5pp, Toolathlon +9.6pp, BFCL-V4 +3.5pp, and SWE-Bench Pro +5.8pp.
  • The trained model adds roughly one quarter as many lines (111.7 versus 415.5) while touching more reference-patch files, indicating more targeted implementation rather than brute-force exploration.
  • Formal-test runs nearly double, from 37.5% to 73.3%, with earlier first-test position, suggesting long-horizon post-training shifts agents toward verification-heavy execution.
  • GDE provides a shared vocabulary for failures across domains: a goal-stability breakdown in a scheduling task and a stale-test override in a repository are the same capability failing at different surfaces.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • If the transfer is causal, it predicts similar gains from other non-software long-horizon domains, such as customer support, scientific data analysis, or healthcare administration, whenever tasks share deep decomposition, parallel synthesis, entangled constraints, and long dependent chains.
  • A natural ablation would hold training data volume constant and vary one task demand at a time; the paper's account predicts that deep decomposition and long dependent chains are the largest drivers because they most directly exercise goal formation and verification.
  • The aggregate metrics suggest a practical test for production pipelines: monitor retrieval repetition, patch size, and test timing as inexpensive proxies for goal-directed execution during post-training, even when the target domain is software engineering.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

4 major / 4 minor

Summary. The manuscript reports a post-training experiment in which Qwen3.5-122B-A10B is trained on 363 Long-Horizon Multi-Tool Agent (LHMTA) office-workflow tasks, with no software-engineering content, and then evaluated on SWE-Bench Pro. The central empirical claim is a pass@1 improvement from 20.5% to 26.3% (+5.8 percentage points) under greedy decoding, which the authors interpret as evidence that long-horizon post-training strengthens a domain-general capability they call goal-directed execution (GDE). GDE is operationalized through four behaviors—goal formation, state construction, goal stability, and verification—and the paper presents paired trajectory analyses and aggregate SWE-Bench Pro metrics as evidence that these behaviors improved after training in both office and software domains. Section 8 explicitly concedes that the experiment is a single model and a single post-training run, that the behavioral framework was refined iteratively on the same trajectories used to illustrate it, that the case analysis is outcome-conditioned, and that no causal or domain-matched counterfactual was run.

Significance. If the +5.8pp transfer is robust, the result is significant: it would show that long-horizon post-training on non-software tasks can improve software-engineering performance, with practical implications for data selection and for behavioral accounts of agent post-training. The paper's strengths are the use of an external benchmark with no fitted parameters, the detailed and deterministic definitions of the aggregate behavioral metrics in Appendix A, and the paired-trajectory methodology that grounds the qualitative analysis in concrete tool calls and artifacts. However, the central transfer claim is a single point estimate with no variance estimate, and the behavioral explanation is partly circular because the four capabilities were developed and demonstrated on the same outcome-conditioned trajectories. The significance of the result therefore hinges on reproducibility and on validation of the behavioral coding that the current manuscript does not yet provide.

major comments (4)
  1. [Section 5.2 and Table 3] The central transfer claim rests on a single post-training run; the paper provides no variance estimate, confidence interval, or significance test for the +5.8pp SWE-Bench Pro gain. LoRA initialization, GSPO rollout sampling, and optimizer noise are stochastic, so training-seed variance could be comparable to or larger than the reported effect. Section 8 concedes this ('one base model and one post-training run'), but it remains load-bearing: without multi-seed replication or at least a reproducibility check, the cross-domain transfer and the GDE interpretation built on it are not yet established.
  2. [Section 6.1 and Section 8] The four GDE capabilities were refined iteratively through qualitative analysis of the same outcome-conditioned trajectories used to demonstrate them, so the claim that 'matched trajectory analysis shows gains in all four GDE behaviors' is partly circular. The framework was not fixed before inspecting the data, and the cases shown were selected for explanatory clarity rather than randomly. An independent, pre-registered coding protocol with blinded annotators and inter-rater reliability, or validation on held-out trajectories, is needed for the behavioral claim to support the transfer account.
  3. [Section 6.3 and Table 5] The aggregate behavioral analysis covers 731 paired SWE-Bench Pro tasks but reports no analysis of the approximately 35 pass-to-fail regressions implied by 77 newly passed tasks and a net gain of 42. The paper's directional narrative—improvement in all four GDE behaviors—is incomplete without understanding whether regressions exhibit the same GDE failures or different ones; this is directly relevant to whether post-training strengthened GDE rather than merely shifted the model's success set.
  4. [Section 5.2] The claimed isolation of training from SWE-Bench Pro is asserted but not documented. Because the transfer interpretation depends on no benchmark instances, graders, checkpoint-selection signal, or hyperparameter-tuning signal from SWE-Bench Pro, the paper should provide a provenance audit or contamination check: dataset hashes, exact task lists, overlap tests with SWE-Bench Pro repositories and issues, and a statement of how the checkpoint was selected. This is a factual premise that cannot be independently verified from the current text.
minor comments (4)
  1. [Abstract and Section 1] There are missing spaces in 'capabilitygoal-directed execution' and 'we usegoal-directed execution'; please fix the typography.
  2. [Section 6.2 and Figures 5-8] The paired figures are illustrative and were selected after observing outcomes; each figure should explicitly state that the cases are not randomly sampled and should include case identifiers or repository links to support independent audit.
  3. [Table 5 and Appendix A] The definition of 'mean first-test position' depends on commands classified as the assistant's shell commands; consider also reporting the fraction of runs with no formal test separately, since normalizing only among tested runs can mask a bimodal distribution.
  4. [Section 2] The related-work discussion mentions HiAgent and ReCAP in a single sentence; a slightly fuller comparison would clarify what the goal-loop representation adds over these existing hierarchical working-memory and recursive-planning approaches.

Circularity Check

1 steps flagged · score 4.0 of 10

Central SWE-Bench transfer is non-circular; the GDE behavioral evidence is partially circular because the four-capability taxonomy was induced from the same outcome-conditioned trajectories it is used to confirm.

  1. fitted input called prediction [Section 6.1 (Behavioral Analysis) and Abstract; cf. Section 8 Limitations]
    "We selected this outcome-conditioned subset to study how newly successful behavior manifested. It does not estimate the prevalence of each capability across the full evaluation distribution. ... Through iterative comparison across cases, we organized recurring behavioral differences into four capabilities: goal formation, state construction, goal stability, and verification. These capabilities form an interpretive framework rather than mutually exclusive labels. [Abstract:] Matched trajectory analysis shows gains in all four GDE behaviors in both office workflows and software repositories."

    The behavioral evidence for GDE is not an independent test of the hypothesis. The 103 analyzed cases were selected precisely because the base model failed and the trained model passed, so behavioral differences are guaranteed by the sampling rule. The four capability labels were then induced from those same cases ('Through iterative comparison across cases, we organized recurring behavioral differences into four capabilities'), and the paper reports 'gains in all four GDE behaviors' as if confirming the framework. This is a post-hoc taxonomy applied to the data that generated it: the reported gains restate the labeling exercise rather than provide independent confirmation.

full rationale

The +5.8pp SWE-Bench Pro improvement is an external, parameter-free benchmark comparison between one base and one post-trained checkpoint; it is not derived from GDE or from any fitted quantity, so the central transfer claim is not circular. The GDE account is partly circular in the specific sense above: the four behaviors were refined on the same outcome-conditioned trajectories used to demonstrate them, so the matched-trajectory 'gains' are a summary of the taxonomy-building process rather than a pre-registered prediction. The paper is transparent about this in Section 8, and no load-bearing self-citation or imported uniqueness theorem appears; the cited prior work from the same group is motivational only. Overall circularity is partial and confined to the interpretive layer, so a score of 4 is appropriate.

Assumptions & free parameters 0 free parameters · 4 assumptions · 2 invented entities

The central transfer result involves no fitted parameters and is a direct benchmark comparison. The GDE explanation, however, rests on several domain assumptions about data isolation, subjective behavioral coding, and proxy validity. The paper introduces two conceptual entities, GDE and the goal loop, both without independent falsifiable evidence outside this paper.

assumptions (4)
  • domain assumption The LHMTA training set contained no software-engineering tasks, graders, or benchmark feedback (Section 5.2).
    This is the load-bearing premise for calling the transfer cross-domain; if false, the result would be same-domain leakage.
  • domain assumption No checkpoint or hyperparameter selection used SWE-Bench Pro or other external benchmark signals (Section 5.2).
    The authors assert this but provide no logs or ablation; it is unverifiable from the paper.
  • domain assumption The four GDE capabilities can be reliably inferred from long trajectories by human review of Claude Opus 4.8 reports (Section 6.1).
    The behavioral account depends on this interpretive coding, which the authors acknowledge is exploratory and involves judgment.
  • domain assumption The aggregate metrics in Table 5 (retrieval spans, patch overlap, added lines, test presence) are valid proxies for GDE behavioral changes (Section 6.3).
    The authors call these indirect signals; reference patches are not unique ground truth, so the metrics could change for reasons unrelated to GDE.
invented entities (2)
  • Goal-directed execution (GDE)
    purpose: Explanatory construct for cross-domain transfer: a set of four behaviors (goal formation, state construction, goal stability, verification) said to be strengthened by long-horizon post-training.
    The construct was operationalized and refined on the same trajectories used to support it; the paper labels the causal link a hypothesis and provides no external falsifiable handle.
  • Goal loop representation
    purpose: A recursive model of task execution (goal, action, working state, verification) used to structure trajectory analysis.
    A descriptive framing borrowed from TOTE and Soar; it is an interpretive lens rather than an independently evidenced mechanism.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Post-Training on Office Work Improves Software Engineering: A Behavioral Account of Cross-Domain Transfer." pith.science (2026). https://pith.science/paper/OHPSZPOP

@misc{pith2026260801604,
  author       = {Pith},
  title        = {Pith review of: Post-Training on Office Work Improves Software Engineering: A Behavioral Account of Cross-Domain Transfer},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/OHPSZPOP}},
  note         = {Machine review of arXiv:2608.01604}
}
read the original abstract

Long-horizon tasks require agents to maintain coherent state and goals across nested and branching work. We call this capability goal-directed execution (GDE): the repeated application of four behaviors, namely selecting goals, constructing task-relevant state, maintaining fidelity to higher-level objectives, and verifying completion against the environment. We hypothesize that long-horizon post-training strengthens these behaviors across domains. We test this by post-training Qwen3.5-122B-A10B on 363 Long-Horizon Multi-Tool Agent (LHMTA) tasks drawn from office workflows. The collection contained no software-engineering tasks, yet the model's pass@1 improved by 5.8 points on SWE-Bench Pro. Matched trajectory analysis shows gains in all four GDE behaviors in both office workflows and software repositories. Aggregate SWE-Bench Pro statistics showed related changes in information gathering, implementation, and verification. Together, the results support a behavioral interpretation in which long-horizon post-training changed how the model organized and applied knowledge across tasks, with effects extending beyond the training domain.

Figures

Figures reproduced from arXiv: 2608.01604 by the authors.

Figure 1
Figure 1. The goal loop. The agent forms a goal, acts on or queries the environment, updates its working [PITH_FULL_IMAGE:figures/full_fig_p004_1.png] view at source ↗
Figure 2
Figure 2. A task becomes a tree of goals. The subscription-review task decomposes into branching subgoals, [PITH_FULL_IMAGE:figures/full_fig_p005_2.png] view at source ↗
Figure 3
Figure 3. Four capabilities within the goal loop: goal formation, state construction, goal stability, and [PITH_FULL_IMAGE:figures/full_fig_p006_3.png] view at source ↗
Figures from the paper (5 more)
Figure 4
Figure 4. Figure 4: Four task demands in LHMTA: deep decomposition, parallel investigation and synthesis, entangled [PITH_FULL_IMAGE:figures/full_fig_p007_4.png]
Figure 5
Figure 5. Figure 5: Goal formation across domains. From the same parent goal and working state, the base model [PITH_FULL_IMAGE:figures/full_fig_p010_5.png]
Figure 6
Figure 6. Figure 6: State construction across domains. The base model drops or misrepresents a relevant observation; [PITH_FULL_IMAGE:figures/full_fig_p011_6.png]
Figure 7
Figure 7. Figure 7: Goal stability across domains. A parent-goal constraint enters a local repair loop; the base model [PITH_FULL_IMAGE:figures/full_fig_p012_7.png]
Figure 8
Figure 8. Figure 8: Verification across domains. The base model relies on proxy or self-referential evidence; the trained [PITH_FULL_IMAGE:figures/full_fig_p012_8.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

49 extracted references · 37 canonical work pages

  1. [1]

    2025 , eprint =

    Arike, Rauno and Donoway, Elizabeth and Bartsch, Henning and Hobbhahn, Marius , title =. 2025 , eprint =. doi:10.48550/arXiv.2505.02709 , url =

    arXiv doi:10.48550/arxiv.2505.02709 2025

  2. [2]

    , title =

    Annett, John and Duncan, Keith D. , title =. Occupational Psychology , volume =

    work page

  3. [3]

    2024 , url =

    Introducing the. 2024 , url =

    work page 2024

  4. [4]

    Proceedings of the National Academy of Sciences , volume =

    Binz, Marcel and Schulz, Eric , title =. Proceedings of the National Academy of Sciences , volume =. 2023 , doi =

    work page 2023

  5. [5]

    , title =

    Bratman, Michael E. , title =

    work page

  6. [6]

    and Moran, Thomas P

    Card, Stuart K. and Moran, Thomas P. and Newell, Allen , title =

    work page

  7. [7]

    Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers) , pages =

    Hu, Mengkang and Chen, Tianxing and Chen, Qiguang and Mu, Yao and Shao, Wenqi and Luo, Ping , title =. Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers) , pages =. 2025 , doi =

    work page 2025

  8. [8]

    2026 , eprint =

    Liu, Zhihan and Guan, Lin and Nie, Yixin and Zhang, Kai and Hao, Zhuoqun and Chen, Lin and Celikyilmaz, Asli and Wang, Zhaoran and Zhang, Na , title =. 2026 , eprint =. doi:10.48550/arXiv.2601.18217 , url =

    work page doi:10.48550/arxiv.2601.18217 2026

Show all 49 references
  1. [9]

    and Dean, Jeff and Devlin, Jacob and Roberts, Adam and Zhou, Denny and Le, Quoc V

    Chung, Hyung Won and Hou, Le and Longpre, Shayne and Zoph, Barret and Tay, Yi and Fedus, William and Li, Yunxuan and Wang, Xuezhi and Dehghani, Mostafa and Brahma, Siddhartha and Webson, Albert and Gu, Shixiang Shane and Dai, Zhuyun and Suzgun, Mirac and Chen, Xinyun and Chowd...

    2024

  2. [10]

    Proceedings of the 37th International Conference on Machine Learning , series =

    Cobbe, Karl and Hesse, Chris and Hilton, Jacob and Schulman, John , title =. Proceedings of the 37th International Conference on Machine Learning , series =. 2020 , url =

    2020

  3. [11]

    2025 , eprint =

    Deng, Xiang and Da, Jeff and Pan, Edwin and He, Yannis Yiming and Ide, Charles and Garg, Kanak and Lauffer, Niklas and Park, Andrew and Pasari, Nitin and Rane, Chetan and Sampath, Karmini and Krishnan, Maya and Kundurthy, Srivatsa and Hendryx, Sean and Wang, Zifan and Bharadwa...

    2025 doi arXiv

  4. [12]

    , title =

    Dietterich, Thomas G. , title =. Journal of Artificial Intelligence Research , volume =. 2000 , doi =

    2000

  5. [13]

    Psychonomic Bulletin & Review , volume =

    Duncan, John and Schramm, Moritz and Thompson, Russell and Dumontheil, Iroise , title =. Psychonomic Bulletin & Review , volume =. 2012 , doi =

    2012

  6. [14]

    Ghallab, Malik and Nau, Dana and Traverso, Paolo , title =

  7. [15]

    Hagendorff, Thilo and Dasgupta, Ishita and Binz, Marcel and Chan, Stephanie C. Y. and Lampinen, Andrew and Wang, Jane X. and Akata, Zeynep and Schulz, Eric , title =. 2023 , eprint =. doi:10.48550/arXiv.2303.13988 , url =

    2023 doi arXiv

  8. [16]

    and Shen, Yelong and Wallis, Phillip and Allen-Zhu, Zeyuan and Li, Yuanzhi and Wang, Shean and Wang, Lu and Chen, Weizhu , title =

    Hu, Edward J. and Shen, Yelong and Wallis, Phillip and Allen-Zhu, Zeyuan and Li, Yuanzhi and Wang, Shean and Wang, Lu and Chen, Weizhu , title =. International Conference on Learning Representations , year =

  9. [17]

    , title =

    Kambhampati, Subbarao and Valmeekam, Karthik and Guan, Lin and Verma, Mudit and Stechly, Kaya and Bhambri, Siddhant and Saldyt, Lucas Paul and Murthy, Anil B. , title =. Proceedings of the 41st International Conference on Machine Learning , series =. 2024 , url =

    2024

  10. [18]

    International Conference on Learning Representations , year =

    Laban, Philippe and Hayashi, Hiroaki and Zhou, Yingbo and Neville, Jennifer , title =. International Conference on Learning Representations , year =

  11. [19]

    and Newell, Allen and Rosenbloom, Paul S

    Laird, John E. and Newell, Allen and Rosenbloom, Paul S. , title =. Artificial Intelligence , volume =. 1987 , doi =

    1987

  12. [20]

    and Pfau, Jacob and Krueger, David , title =

    Langosco, Lauro Langosco Di and Koch, Jack and Sharkey, Lee D. and Pfau, Jacob and Krueger, David , title =. Proceedings of the 39th International Conference on Machine Learning , series =. 2022 , url =

    2022

  13. [21]

    International Conference on Learning Representations , year =

    Li, Junlong and Zhao, Wenshuo and Zhao, Jian and Zeng, Weihao and Wu, Haoze and Wang, Xiaochen and Ge, Rui and Cao, Yuxuan and Huang, Yuzhen and Liu, Wei and Liu, Junteng and Su, Zhaochen and Guo, Yiyang and Zhou, Fan and Zhang, Lueyang and Michelini, Juan and Wang, Xingyao an...

  14. [22]

    International Conference on Learning Representations , year =

    Liu, Wei and Zeng, Weihao and He, Keqing and Jiang, Yong and He, Junxian , title =. International Conference on Learning Representations , year =

  15. [23]

    2026 , eprint =

    Mehta, Sushant and Ritchie, Logan and Garre, Suhaas and Niebres, Ian and Heiner, Nick and Chen, Edwin , title =. 2026 , eprint =. doi:10.48550/arXiv.2602.16179 , url =

    2026 doi

  16. [24]

    and Galanter, Eugene and Pribram, Karl H

    Miller, George A. and Galanter, Eugene and Pribram, Karl H. , title =

  17. [25]

    and Cohen, Jonathan D

    Miller, Earl K. and Cohen, Jonathan D. , title =. Annual Review of Neuroscience , volume =. 2001 , doi =

    2001

  18. [26]

    , title =

    Newell, Allen and Simon, Herbert A. , title =

  19. [27]

    and Leike, Jan and Lowe, Ryan , title =

    Ouyang, Long and Wu, Jeffrey and Jiang, Xu and Almeida, Diogo and Wainwright, Carroll and Mishkin, Pamela and Zhang, Chong and Agarwal, Sandhini and Slama, Katarina and Ray, Alex and Schulman, John and Hilton, Jacob and Kelton, Fraser and Miller, Luke and Simens, Maddie and As...

    2022

  20. [28]

    and Mao, Huanzhi and Yan, Fanjia and Ji, Charlie Cheng-Jie and Suresh, Vishnu and Stoica, Ion and Gonzalez, Joseph E

    Patil, Shishir G. and Mao, Huanzhi and Yan, Fanjia and Ji, Charlie Cheng-Jie and Suresh, Vishnu and Stoica, Ion and Gonzalez, Joseph E. , title =. Proceedings of the 42nd International Conference on Machine Learning , series =. 2025 , url =

    2025

  21. [29]

    Findings of the Association for Computational Linguistics: NAACL 2024 , publisher =

    Prasad, Archiki and Koller, Alexander and Hartmann, Mareike and Clark, Peter and Sabharwal, Ashish and Bansal, Mohit and Khot, Tushar , title =. Findings of the Association for Computational Linguistics: NAACL 2024 , publisher =. 2024 , doi =

    2024

  22. [30]

    Machine behaviour , journal =

    Rahwan, Iyad and Cebrian, Manuel and Obradovich, Nick and Bongard, Josh and Bonnefon, Jean-Fran. Machine behaviour , journal =. 2019 , doi =

    2019

  23. [31]

    and Georgeff, Michael P

    Rao, Anand S. and Georgeff, Michael P. , title =. Proceedings of the First International Conference on Multi-Agent Systems , editor =. 1995 , isbn =

    1995

  24. [32]

    2026 , eprint =

    Ritchie, Logan and Mehta, Sushant and Heiner, Nick and Yu, Mason and Chen, Edwin , title =. 2026 , eprint =. doi:10.48550/arXiv.2601.09032 , url =

    2026 doi

  25. [33]

    2022 , eprint =

    Shah, Rohin and Varma, Vikrant and Kumar, Ramana and Phuong, Mary and Krakovna, Victoria and Uesato, Jonathan and Kenton, Zac , title =. 2022 , eprint =. doi:10.48550/arXiv.2210.01790 , url =

    2022 doi arXiv

  26. [34]

    Advances in Neural Information Processing Systems , volume =

    Shinn, Noah and Cassano, Federico and Gopinath, Ashwin and Narasimhan, Karthik and Yao, Shunyu , title =. Advances in Neural Information Processing Systems , volume =. 2023 , doi =

    2023

  27. [35]

    and Yao, Shunyu and Narasimhan, Karthik and Griffiths, Thomas L

    Sumers, Theodore R. and Yao, Shunyu and Narasimhan, Karthik and Griffiths, Thomas L. , title =. Transactions on Machine Learning Research , year =

  28. [36]

    Proceedings of the Fifth Workshop on Generation, Evaluation and Metrics (

    Mehta, Sushant and Panavas, Liudas and Fleming, Eleanor and Mains, Paul and Chen, Edwin , title =. Proceedings of the Fifth Workshop on Generation, Evaluation and Metrics (. 2026 , doi =

    2026

  29. [37]

    and Precup, Doina and Singh, Satinder , title =

    Sutton, Richard S. and Precup, Doina and Singh, Satinder , title =. Artificial Intelligence , volume =. 1999 , doi =

    1999

  30. [38]

    , title =

    Sutton, Richard S. , title =. 2019 , url =

    2019

  31. [39]

    Advances in Neural Information Processing Systems , volume =

    Valmeekam, Karthik and Marquez, Matthew and Olmo, Alberto and Sreedharan, Sarath and Kambhampati, Subbarao , title =. Advances in Neural Information Processing Systems , volume =. 2023 , doi =

    2023

  32. [40]

    and Guu, Kelvin and Yu, Adams Wei and Lester, Brian and Du, Nan and Dai, Andrew M

    Wei, Jason and Bosma, Maarten and Zhao, Vincent Y. and Guu, Kelvin and Yu, Adams Wei and Lester, Brian and Du, Nan and Dai, Andrew M. and Le, Quoc V. , title =. International Conference on Learning Representations , year =

  33. [41]

    Proceedings of the 41st International Conference on Machine Learning , series =

    Xia, Mengzhou and Malladi, Sadhika and Gururangan, Suchin and Arora, Sanjeev and Chen, Danqi , title =. Proceedings of the 41st International Conference on Machine Learning , series =. 2024 , url =

    2024

  34. [42]

    International Conference on Learning Representations , year =

    Yao, Shunyu and Zhao, Jeffrey and Yu, Dian and Du, Nan and Shafran, Izhak and Narasimhan, Karthik and Cao, Yuan , title =. International Conference on Learning Representations , year =

  35. [43]

    and Cao, Yuan and Narasimhan, Karthik , title =

    Yao, Shunyu and Yu, Dian and Zhao, Jeffrey and Shafran, Izhak and Griffiths, Thomas L. and Cao, Yuan and Narasimhan, Karthik , title =. Advances in Neural Information Processing Systems , volume =. 2023 , doi =

    2023

  36. [44]

    Instruction Diversity Drives Generalization to Unseen Tasks , year =

    Zhang, Dylan and Wang, Justin and Charton, Fran. Instruction Diversity Drives Generalization to Unseen Tasks , year =. doi:10.48550/arXiv.2402.10891 , url =. 2402.10891 , archiveprefix =

    doi arXiv

  37. [45]

    Advances in Neural Information Processing Systems , volume =

    Zhang, Zhenyu and Chen, Tianyi and Xu, Weiran and Pentland, Alex and Pei, Jiaxin , title =. Advances in Neural Information Processing Systems , volume =. 2025 , url =

    2025

  38. [46]

    2025 , eprint =

    Zheng, Chujie and Liu, Shixuan and Li, Mingze and Chen, Xiong-Hui and Yu, Bowen and Gao, Chang and Dang, Kai and Liu, Yuqiong and Men, Rui and Yang, An and Zhou, Jingren and Lin, Junyang , title =. 2025 , eprint =. doi:10.48550/arXiv.2507.18071 , url =

    2025 doi arXiv

  39. [47]

    Least-to-Most Prompting Enables Complex Reasoning in Large Language Models , booktitle =

    Zhou, Denny and Sch. Least-to-Most Prompting Enables Complex Reasoning in Large Language Models , booktitle =. 2023 , url =

    2023

  40. [48]

    Advances in Neural Information Processing Systems , volume =

    Zhou, Chunting and Liu, Pengfei and Xu, Puxin and Iyer, Srinivasan and Sun, Jiao and Mao, Yuning and Ma, Xuezhe and Efrat, Avia and Yu, Ping and Yu, Lili and Zhang, Susan and Ghosh, Gargi and Lewis, Mike and Zettlemoyer, Luke and Levy, Omer , title =. Advances in Neural Inform...

    2023

  41. [49]

    Proceedings of the 41st International Conference on Machine Learning , series =

    Zhou, Andy and Yan, Kai and Shlapentokh-Rothman, Michal and Wang, Haohan and Wang, Yu-Xiong , title =. Proceedings of the 41st International Conference on Machine Learning , series =. 2024 , url =

    2024

Pith tools

Reviewed August 15, 2026 · model on record in the stance chip above.