Pith. sign in

REVIEW 2 major objections 1 minor 2 cited by

Shepherd makes an agent's full execution a first-class reversible object that meta-agents can inspect and edit.

Reviewed by Pith at T0; open to challenge. T0 means a machine referee read the full paper against a public rubric. the ladder, T0–T4 →

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

T0 review · grok-4.3

2026-06-30 22:22 UTC pith:YQHAX3G6

load-bearing objection Shepherd gives meta-agents direct access to reversible execution traces, which is a practical step forward, but the experiments leave open whether those traces actually contain everything needed without extra context. the 2 major comments →

arxiv 2605.10913 v3 pith:YQHAX3G6 submitted 2026-05-11 cs.AI cs.PLcs.SE

Shepherd: Enabling Programmable Meta-Agents via Reversible Agentic Execution Traces

Simon Yu , Derek Chong , Ananjan Nandi , Dilara Soylu , Jiuding Sun , Christopher D Manning , Weiyan Shi This is my paper
classification cs.AI cs.PLcs.SE
keywords meta-agentsagentic execution tracesreversible tracesLLM agentssupervisioncounterfactual optimizationagentic RL
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

The pith

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

The paper introduces Shepherd, a Python substrate that records every model action, tool call, and environment change as a structured event inside a Git-like trace. Because the trace is reversible and first-class, meta-agents can halt runs, propose edits, and replay from chosen points without rebuilding state from transcripts or snapshots. The approach is demonstrated on three tasks: conflict prevention among parallel coding agents, workflow repair through counterfactual edits, and improved credit assignment during agentic reinforcement learning.

Core claim

Shepherd grounds agent execution in functional programming principles so that the entire run becomes a manipulable object. Every action becomes a structured event stored in a reversible, Git-like trace that supports fast reversion to any prior state. This substrate lets meta-agents perform inspection, transformation, halting, and repair operations directly on the execution history rather than through ad-hoc reconstruction.

What carries the argument

The reversible Git-like execution trace that turns every model action, tool call, and state change into a structured, first-class event.

Load-bearing premise

The structured events captured in the trace contain all state and context that meta-agents need to inspect, transform, halt, or repair runs without losing critical information.

What would settle it

A concrete workflow in which a meta-agent given only the Shepherd trace cannot correctly identify or repair the cause of failure because some required context is absent from the captured events.

Watch this falsifier — get emailed when new claim-graph text bears on it.

If this is right

  • A supervisor meta-agent can detect and prevent conflicts among parallel coding agents, raising pair-coding pass rate from 28.8% to 54.7% on CooperBench.
  • A counterfactual optimization meta-agent can propose edits and replay from the point of changed behavior, outperforming MetaHarness on Terminal-Bench 2.0 by 12.8% with 58% lower wall-clock time.
  • A training meta-agent can select fork points during rollouts to improve credit assignment, doubling GRPO's uplift on Terminal-Bench 2.0.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • The same trace format could let meta-agents coordinate agents written in different base frameworks if those frameworks adopt compatible event structures.
  • Version-control operations such as branching execution paths or merging successful sub-runs become natural extensions once traces are first-class objects.
  • Debugging and auditing tools for agent systems could shift from post-hoc log analysis to direct manipulation of the live reversible history.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

2 major / 1 minor

Summary. The manuscript introduces Shepherd, a Python substrate grounded in functional programming principles that makes an agent's execution a first-class reversible object. Every model action, tool call, and environment change is logged as a structured event in a Git-like trace that supports inspection, transformation, halting, and repair by meta-agents without ad-hoc tooling. Three use cases are presented: a supervisor meta-agent for parallel coding agents (CooperBench pass rate lifted from 28.8% to 54.7%), a counterfactual optimization meta-agent (outperforming MetaHarness on Terminal-Bench 2.0 by 12.8% with 58% lower wall-clock), and a training meta-agent for credit assignment in agentic RL (doubling GRPO uplift on Terminal-Bench 2.0). The system is open-sourced.

Significance. If the reversible trace mechanism fully captures execution state, the substrate could provide a principled foundation for meta-agent operations that current transcript- or snapshot-based systems lack. The reported speed advantage over docker commit/fork and the open-source release are concrete strengths that would aid reproducibility and further experimentation in agentic systems.

major comments (2)
  1. [Abstract / use cases] Abstract and use-case descriptions: the central claim that structured events in the reversible trace contain all necessary state and context for meta-agent inspection, transformation, halting, and repair (without ad-hoc tooling) is load-bearing, yet the manuscript supplies no explicit verification or ablation showing that the three meta-agent implementations operated exclusively from the declared event types rather than reconstructing missing mutable state, globals, or side effects.
  2. [Abstract] Performance reporting (Abstract): the reported gains (28.8%→54.7% on CooperBench; 12.8% and 58% wall-clock on Terminal-Bench 2.0; doubling GRPO uplift) are presented without baseline definitions, error bars, run counts, or data-exclusion rules, preventing assessment of whether the trace sufficiency assumption holds or whether gains could be reproduced from the trace alone.
minor comments (1)
  1. [Abstract] The 5× speedup claim versus docker commit/fork would benefit from a precise experimental setup (hardware, trace sizes, commit frequency) to allow direct comparison.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive comments, which help clarify the presentation of Shepherd's core claims and experimental reporting. We address each major comment below and indicate where revisions will be made.

read point-by-point responses

  1. Referee: [Abstract / use cases] Abstract and use-case descriptions: the central claim that structured events in the reversible trace contain all necessary state and context for meta-agent inspection, transformation, halting, and repair (without ad-hoc tooling) is load-bearing, yet the manuscript supplies no explicit verification or ablation showing that the three meta-agent implementations operated exclusively from the declared event types rather than reconstructing missing mutable state, globals, or side effects.

    Authors: We agree that an explicit verification or ablation would strengthen the load-bearing claim. The manuscript's design rests on the functional programming substrate ensuring that every state change is captured as a structured event, with no hidden mutable globals or side effects outside the trace (as described in Sections 3 and 4). However, the use-case implementations do not include a dedicated ablation confirming exclusive reliance on these events. We will revise the manuscript to add a verification subsection (likely in Section 5 or an appendix) that documents the event types used by each meta-agent and confirms no external state reconstruction occurred. revision: yes

  2. Referee: [Abstract] Performance reporting (Abstract): the reported gains (28.8%→54.7% on CooperBench; 12.8% and 58% wall-clock on Terminal-Bench 2.0; doubling GRPO uplift) are presented without baseline definitions, error bars, run counts, or data-exclusion rules, preventing assessment of whether the trace sufficiency assumption holds or whether gains could be reproduced from the trace alone.

    Authors: We acknowledge that the abstract and associated results sections lack the requested details on baseline definitions, error bars, number of runs, and data-exclusion criteria. These omissions limit reproducibility assessment. We will revise the abstract, results tables, and experimental sections to include precise baseline descriptions, statistical reporting (including run counts and variance where available), and any applicable exclusion rules from our evaluation protocol. revision: yes

Circularity Check

0 steps flagged

No circularity: system description with independent empirical use cases

full rationale

The paper presents Shepherd as a new Python substrate for agent execution traces, with three use-case demonstrations of performance gains on external benchmarks (CooperBench, Terminal-Bench 2.0). No equations, fitted parameters, or derivations appear in the provided text. Claims rest on the system's design and measured outcomes rather than any self-definition, renamed known result, or load-bearing self-citation chain. The central premise that structured events suffice for meta-agent operations is an engineering assumption tested empirically, not a reduction to inputs by construction.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

Abstract-only review provides no information on free parameters, axioms, or invented entities; none can be identified.

pith-pipeline@v0.9.1-grok · 5850 in / 1106 out tokens · 30157 ms · 2026-06-30T22:22:33.821749+00:00 · methodology

0 comments
read the original abstract

As LLM agent systems take on more complex tasks, they increasingly rely on meta-agents: higher-order agents that create, operate on and manage other agents. Meta-agent operations such as coordinating agents, halting risky actions before execution, or repairing failed runs, require runtime manipulation of agentic execution. Yet existing agentic substrates make this difficult: they expose only transcripts and environment snapshots, forcing meta-agents to build ad hoc tooling to reconstruct and operate over full execution state. Therefore, we introduce Shepherd, a Python substrate grounded in functional programming principles, where an agent's execution is itself a first-class object that a meta-agent can easily inspect and transform. Every model action, tool call, and environment change becomes a structured event in a reversible, Git-like execution trace, where any past state can be reverted 5x faster than docker commit and fork. Three example use cases show Shepherd's versatility: (1) a supervisor meta-agent prevents conflicts among parallel coding agents, lifting pair-coding pass rate from 28.8% to 54.7% on CooperBench; (2) a counterfactual optimization meta-agent repairs agent workflows by proposing edits and replaying runs from the point of changed behavior, outperforming MetaHarness on Terminal-Bench 2.0 by 12.8% with 58% lower wall-clock; (3) a training meta-agent picks fork points during rollouts to improve credit assignment in long-horizon agentic RL, doubling GRPO's uplift on Terminal-Bench 2.0. We open-source Shepherd to enable principled and efficient operations over agentic execution for both users and meta-agents.

Figures

Figures reproduced from arXiv: 2605.10913 by Ananjan Nandi, Christopher D Manning, Derek Chong, Dilara Soylu, Jiuding Sun, Simon Yu, Weiyan Shi.

Figure 1
Figure 1. Figure 1: SHEPHERD meta-agents. Top: A supervisor meta-agent manages code repair agents. Bottom: Results from three meta-agents: (A) live supervision; (B) meta-optimization; (C) Tree GRPO Abstract As LLM-based agentic systems grow more complex, they increasingly rely on meta-agents: higher-order agents that act on other agents, much like managers supervise employees. Yet existing agentic runtimes expose execution on… view at source ↗
Figure 2
Figure 2. Figure 2: Live intervention experiments on CooperBench, with Claude Haiku 4.5 as worker. [PITH_FULL_IMAGE:figures/full_fig_p007_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: LiveCodeBench comparison. Left: held-out test pass-rate versus optimization wallclock. Right: dev-set trajectory for each method across optimization wallclock. CRO subtask-cache reuse is reported separately in [PITH_FULL_IMAGE:figures/full_fig_p008_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: CRO computation reuse on LiveCodeBench rises from ∼1% on the first cold proposer session to over 60%. Setup. We evaluate on subsets of HoVer [13], MATH [8], IFBench [27], LiveCodeBench [11], and Terminal￾Bench 2.0 (TB-2; [24]), comparing CRO against the base￾line workflow, GEPA (optimizing workflow code) [2], and MetaHarness [19]. The executor is GPT-5.4-mini and meta-optimizers use GPT-5.4 (in the Codex h… view at source ↗
Figure 4
Figure 4. Figure 4: LiveCodeBench comparison. Left: held-out test pass-rate versus optimization wall-clock. Right: dev-set trajectory for each method across optimization wall-clock. CRO subtask-cache reuse is reported separately in [PITH_FULL_IMAGE:figures/full_fig_p009_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Trajectory compression across two worker model families and two benchmarks. The same [PITH_FULL_IMAGE:figures/full_fig_p020_5.png] view at source ↗
Figure 5
Figure 5. Figure 5: Computation reuse on Live￾CodeBench with CRO. We attribute CRO’s downstream performance gains to two sources. First, CRO’s counterfactual experiments hold the unaffected part of the computation constant, letting the model isolate and patch failures, compared to the noise introduced by MetaHarness’s full-pipeline reruns; we find qualitatively that the generated hypotheses are high-quality (Appendix F). Seco… view at source ↗
Figure 6
Figure 6. Figure 6: HoVer: held-out test pass-rate vs. optimization wallclock (left) and per-iteration dev-set [PITH_FULL_IMAGE:figures/full_fig_p039_6.png] view at source ↗
Figure 6
Figure 6. Figure 6: Trajectory compression across two worker model families and two benchmarks. The same [PITH_FULL_IMAGE:figures/full_fig_p022_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: HoVer: subtask cache reuse per CRO proposer session. [PITH_FULL_IMAGE:figures/full_fig_p040_7.png] view at source ↗
Figure 8
Figure 8. Figure 8: IFBench: test pass-rate vs. wallclock and dev-set trajectory. [PITH_FULL_IMAGE:figures/full_fig_p040_8.png] view at source ↗
Figure 8
Figure 8. Figure 8: HoVer: subtask cache reuse per CRO proposer session. [PITH_FULL_IMAGE:figures/full_fig_p041_8.png] view at source ↗
Figure 9
Figure 9. Figure 9: IFBench: subtask cache reuse per CRO proposer session. [PITH_FULL_IMAGE:figures/full_fig_p040_9.png] view at source ↗
Figure 9
Figure 9. Figure 9: IFBench: test pass-rate vs. wall-clock and dev-set trajectory. [PITH_FULL_IMAGE:figures/full_fig_p041_9.png] view at source ↗
Figure 10
Figure 10. Figure 10: LiveCodeBench: test pass-rate vs. wallclock and dev-set trajectory. [PITH_FULL_IMAGE:figures/full_fig_p041_10.png] view at source ↗
Figure 11
Figure 11. Figure 11: LiveCodeBench: subtask cache reuse per CRO proposer session. [PITH_FULL_IMAGE:figures/full_fig_p041_11.png] view at source ↗
Figure 11
Figure 11. Figure 11: LiveCodeBench: test pass-rate vs. wall-clock and dev-set trajectory. [PITH_FULL_IMAGE:figures/full_fig_p042_11.png] view at source ↗
Figure 12
Figure 12. Figure 12: MATH (Level 5): test pass-rate vs. wallclock and dev-set trajectory. [PITH_FULL_IMAGE:figures/full_fig_p042_12.png] view at source ↗
Figure 13
Figure 13. Figure 13: MATH (Level 5): subtask cache reuse per CRO proposer session. [PITH_FULL_IMAGE:figures/full_fig_p042_13.png] view at source ↗
Figure 13
Figure 13. Figure 13: MATH (Level 5): test pass-rate vs. wall-clock and dev-set trajectory. [PITH_FULL_IMAGE:figures/full_fig_p043_13.png] view at source ↗
Figure 14
Figure 14. Figure 14: TerminalBench 2.0: test pass-rate vs. wallclock and dev-set trajectory. [PITH_FULL_IMAGE:figures/full_fig_p042_14.png] view at source ↗
Figure 14
Figure 14. Figure 14: MATH (Level 5): subtask cache reuse per CRO proposer session. [PITH_FULL_IMAGE:figures/full_fig_p043_14.png] view at source ↗
Figure 15
Figure 15. Figure 15: TerminalBench 2.0: subtask cache reuse per CRO proposer session. [PITH_FULL_IMAGE:figures/full_fig_p043_15.png] view at source ↗
Figure 16
Figure 16. Figure 16: GRPO group composition over training (rows: base model; columns: setting). Tree-GRPO [PITH_FULL_IMAGE:figures/full_fig_p047_16.png] view at source ↗
Figure 16
Figure 16. Figure 16: TerminalBench 2.0: subtask cache reuse per CRO proposer session. [PITH_FULL_IMAGE:figures/full_fig_p044_16.png] view at source ↗
Figure 17
Figure 17. Figure 17: Held-out Endless Terminals evaluation, sampled every 10 training steps (raw, unsmoothed). [PITH_FULL_IMAGE:figures/full_fig_p047_17.png] view at source ↗
Figure 17
Figure 17. Figure 17: GRPO group composition over training (rows: base model; columns: setting). Tree-GRPO [PITH_FULL_IMAGE:figures/full_fig_p048_17.png] view at source ↗
Figure 18
Figure 18. Figure 18: Train raw reward (mean over G=8 roots) for both base models, panels are Qwen3.5- 35B-A3B (left) and Nemotron-3-Super-120B-A12B (right). Tree-GRPO (K=4, teal) reaches higher reward than Flat GRPO (red) at every rollout step. Faint dots are observed steps from the flat-baseline run; smooth lines are denoised trajectories. Case 1: Early mistake (T=4, reward=0.00) Task: Install the requests package and verify… view at source ↗
Figure 19
Figure 19. Figure 19: Early-mistake case. The wrong package name on turn 1 dooms the rest of the trajectory. [PITH_FULL_IMAGE:figures/full_fig_p048_19.png] view at source ↗
Figure 19
Figure 19. Figure 19: Train raw reward (mean over G=8 roots) for both base models, panels are Qwen3.5- 35B-A3B (left) and Nemotron-3-Super-120B-A12B (right). Tree-GRPO (K=4, teal) reaches higher reward than Flat GRPO (red) at every rollout step. Faint dots are observed steps from the flat-baseline run; smooth lines are denoised trajectories. Case 1: Early mistake (T=4, reward=0.00) Task: Install the requests package and verify… view at source ↗
Figure 20
Figure 20. Figure 20: Ambiguous case. At least three turns offer plausible branches (skip-the-version-check, [PITH_FULL_IMAGE:figures/full_fig_p049_20.png] view at source ↗
Figure 21
Figure 21. Figure 21: Long-trajectory case. A 9-turn rollout with a wrong-file edit at turn 4 cascades into 5 [PITH_FULL_IMAGE:figures/full_fig_p049_21.png] view at source ↗
Figure 21
Figure 21. Figure 21: Ambiguous case. At least three turns offer plausible branches (skip-the-version-check, [PITH_FULL_IMAGE:figures/full_fig_p050_21.png] view at source ↗
Figure 22
Figure 22. Figure 22: Long-trajectory case. A 9-turn rollout with a wrong-file edit at turn 4 cascades into 5 [PITH_FULL_IMAGE:figures/full_fig_p050_22.png] view at source ↗

Review history (2 revisions) →

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Forward citations

Cited by 2 Pith papers

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score.

  1. From Cognitive Architectures to Language Agents: A Mechanism-Level Review of Lineage, Convergence, and Migration Gaps

    cs.AI 2026-07 conditional novelty 6.0

    Coding each mechanism for evidence of lineage and implementation depth, the review closes one candidate gap (GraSP) and isolates five residual control bundles for language agents.

  2. Claim Plane: Enforceable Change Intents and Dynamic Scope for Parallel Coding Agents

    cs.SE 2026-07 accept novelty 6.0

    Claim Plane deterministically admits versioned change intents and promotes contingent scope just in time, so parallel coding agents can only mutate what they are currently authorized to touch.

Reference graph

Works this paper leans on

83 extracted references · 83 canonical work pages · cited by 2 Pith papers · 28 internal anchors

  1. [1]

    URL https://www.anthropic

    https://www.anthropic.com/engineering/managed-agents. URL https://www.anthropic. com/engineering/managed-agents

    work page

  2. [2]

    GEPA: Reflective Prompt Evolution Can Outperform Reinforcement Learning

    Lakshya A. Agrawal, Shangyin Tan, Dilara Soylu, Noah Ziems, Rishi Khare, Krista Opsahl- Ong, Arnav Singhvi, Herumb Shandilya, Michael J. Ryan, Meng Jiang, Christopher Potts, Koushik Sen, Alexandros G. Dimakis, Ion Stoica, Dan Klein, Matei Zaharia, and Omar Khattab. GEPA: Reflective Prompt Evolution Can Outperform Reinforcement Learning, July 2025. URL htt...

    work page internal anchor Pith review Pith/arXiv arXiv 2025

  3. [3]

    How to Train Your Advisor: Steering Black-Box LLMs with Advisor Models

    Parth Asawa, Alan Zhu, Abby O’Neill, Matei Zaharia, Alexandros G. Dimakis, and Joseph E. Gonzalez. How to Train Your Advisor: Steering Black-Box LLMs with Advisor Models, October 2025. URLhttp://arxiv.org/abs/2510.02453. arXiv:2510.02453 [cs]

    work page internal anchor Pith review Pith/arXiv arXiv 2025

  4. [4]

    Why Do Multi-Agent LLM Systems Fail?

    Mert Cemri, Melissa Z. Pan, Shuyi Yang, Lakshya A. Agrawal, Bhavya Chopra, Rishabh Tiwari, Kurt Keutzer, Aditya Parameswaran, Dan Klein, Kannan Ramchandran, Matei Zaharia, Joseph E. Gonzalez, and Ion Stoica. Why do multi-agent llm systems fail?, 2025. URL https://arxiv.org/abs/2503.13657

    work page internal anchor Pith review Pith/arXiv arXiv 2025

  5. [5]

    Context-lite multi-turn reinforcement learning for LLM agents

    Wentse Chen, Jiayu Chen, Hao Zhu, and Jeff Schneider. Context-lite multi-turn reinforcement learning for LLM agents. InES-FoMo III: 3rd Workshop on Efficient Systems for Foundation Models, 2025. URLhttps://openreview.net/forum?id=6CE5PLsZdW

    work page 2025

  6. [6]

    Advances in Neural Information Processing Systems (NeurIPS) , year=

    Ching-An Cheng, Allen Nie, and Adith Swaminathan. Trace is the Next AutoDiff: Generative Optimization with Rich Feedback, Execution Traces, and LLMs, June 2024. URL https: //arxiv.org/abs/2406.16218. arXiv:2406.16218

    work page arXiv 2024

  7. [7]

    Endless terminals: Scaling rl environments for terminal agents.arXiv preprint arXiv:2601.16443,

    Kanishk Gandhi, Shivam Garg, Noah D. Goodman, and Dimitris Papailiopoulos. Endless Terminals: Scaling RL Environments for Terminal Agents, January 2026. URL https:// arxiv.org/abs/2601.16443. arXiv:2601.16443

    work page arXiv 2026

  8. [8]

    Measuring mathematical problem solving with the math dataset,

    Dan Hendrycks, Collin Burns, Saurav Kadavath, Akul Arora, Steven Basart, Eric Tang, Dawn Song, and Jacob Steinhardt. Measuring mathematical problem solving with the math dataset,

    work page

  9. [9]

    URLhttps://arxiv.org/abs/2103.03874

    work page internal anchor Pith review Pith/arXiv arXiv

  10. [10]

    MetaGPT: Meta Programming for A Multi-Agent Collaborative Framework

    Sirui Hong, Mingchen Zhuge, Jiaqi Chen, Xiawu Zheng, Yuheng Cheng, Ceyao Zhang, Jinlin Wang, Zili Wang, Steven Ka Shing Yau, Zijuan Lin, Liyang Zhou, Chenyu Ran, Lingfeng Xiao, Chenglin Wu, and Jürgen Schmidhuber. MetaGPT: Meta Programming for A Multi- Agent Collaborative Framework, August 2023. URLhttps://arxiv.org/abs/2308.00352. arXiv:2308.00352

    work page internal anchor Pith review Pith/arXiv arXiv 2023

  11. [11]

    Treerl: Llm rein- forcement learning with on-policy tree search, 2025

    Zhenyu Hou, Ziniu Hu, Yujiang Li, Rui Lu, Jie Tang, and Yuxiao Dong. Treerl: Llm rein- forcement learning with on-policy tree search, 2025. URL https://arxiv.org/abs/2506. 11902

    work page 2025

  12. [12]

    LiveCodeBench: Holistic and Contamination Free Evaluation of Large Language Models for Code

    Naman Jain, King Han, Alex Gu, Wen-Ding Li, Fanjia Yan, Tianjun Zhang, Sida Wang, Ar- mando Solar-Lezama, Koushik Sen, and Ion Stoica. Livecodebench: Holistic and contamination free evaluation of large language models for code, 2024. URL https://arxiv.org/abs/ 2403.07974

    work page internal anchor Pith review Pith/arXiv arXiv 2024

  13. [13]

    Tree search for llm agent reinforcement learning.arXiv preprint arXiv:2509.21240, 2025

    Yuxiang Ji, Ziyu Ma, Yong Wang, Guanhua Chen, Xiangxiang Chu, and Liaoni Wu. Tree search for llm agent reinforcement learning, 2026. URLhttps://arxiv.org/abs/2509.21240

    work page arXiv 2026

  14. [14]

    Hover: A dataset for many-hop fact extraction and claim verification, 2020

    Yichen Jiang, Shikha Bordia, Zheng Zhong, Charles Dognin, Maneesh Singh, and Mohit Bansal. Hover: A dataset for many-hop fact extraction and claim verification, 2020. URL https://arxiv.org/abs/2011.03088. 11

    work page arXiv 2020

  15. [15]

    SWE-bench: Can language models resolve real-world github issues? InThe Twelfth International Conference on Learning Representations, 2024

    Carlos E Jimenez, John Yang, Alexander Wettig, Shunyu Yao, Kexin Pei, Ofir Press, and Karthik R Narasimhan. SWE-bench: Can language models resolve real-world github issues? InThe Twelfth International Conference on Learning Representations, 2024. URL https: //openreview.net/forum?id=VTF8yNQM66

    work page 2024

  16. [16]

    DSPy: Compiling Declarative Language Model Calls into Self-Improving Pipelines

    Omar Khattab, Arnav Singhvi, Paridhi Maheshwari, Zhiyuan Zhang, Keshav Santhanam, Sri Vardhamanan, Saiful Haq, Ashutosh Sharma, Thomas T. Joshi, Hanna Moazam, Heather Miller, Matei Zaharia, and Christopher Potts. DSPy: Compiling Declarative Language Model Calls into Self-Improving Pipelines, October 2023. URLhttps://arxiv.org/abs/2310.03714. arXiv:2310.03714

    work page internal anchor Pith review Pith/arXiv arXiv 2023

  17. [17]

    arXiv preprint arXiv:2601.13295 , year=

    Arpandeep Khatua, Hao Zhu, Peter Tran, Arya Prabhudesai, Frederic Sadrieh, Johann K. Lieberwirth, Xinkai Yu, Yicheng Fu, Michael J. Ryan, Jiaxin Pei, and Diyi Yang. CooperBench: Why Coding Agents Cannot be Your Teammates Yet, January 2026. URL https://arxiv. org/abs/2601.13295. arXiv:2601.13295

    work page arXiv 2026

  18. [18]

    Scaling Test-Time Compute for Agentic Coding

    Joongwon Kim, Wannan Yang, Kelvin Niu, Hongming Zhang, Yun Zhu, Eryk Helenowski, Ruan Silva, Zhengxing Chen, Srinivasan Iyer, Manzil Zaheer, Daniel Fried, Hannaneh Hajishirzi, Sanjeev Arora, Gabriel Synnaeve, Ruslan Salakhutdinov, and Anirudh Goyal. Scaling Test-Time Compute for Agentic Coding, 2026. URLhttps://arxiv.org/abs/2604.16529

    work page internal anchor Pith review Pith/arXiv arXiv 2026

  19. [19]

    Tinker, 2025

    Thinking Machines Lab. Tinker, 2025. URLhttps://thinkingmachines.ai/tinker/

    work page 2025

  20. [20]

    Meta-Harness: End-to-End Optimization of Model Harnesses

    Yoonho Lee, Roshen Nair, Qizheng Zhang, Kangwook Lee, Omar Khattab, and Chelsea Finn. Meta-Harness: End-to-End Optimization of Model Harnesses, March 2026. URL https://arxiv.org/abs/2603.28052. arXiv:2603.28052

    work page internal anchor Pith review Pith/arXiv arXiv 2026

  21. [21]

    Agentic Aggregation for Parallel Scaling of Long-Horizon Agentic Tasks

    Yoonsang Lee, Howard Yen, Xi Ye, and Danqi Chen. Agentic Aggregation for Parallel Scaling of Long-Horizon Agentic Tasks, 2026. URLhttps://arxiv.org/abs/2604.11753

    work page internal anchor Pith review Pith/arXiv arXiv 2026

  22. [22]

    Combee: Scaling Prompt Learning for Self-Improving Language Model Agents

    Hanchen Li, Runyuan He, Qizheng Zhang, Changxiu Ji, Qiuyang Mang, Xiaokun Chen, Lakshya A. Agrawal, Wei-Liang Liao, Eric Yang, Alvin Cheung, James Zou, Kunle Oluko- tun, Ion Stoica, and Joseph E. Gonzalez. Combee: Scaling Prompt Learning for Self- Improving Language Model Agents, April 2026. URLhttp://arxiv.org/abs/2604.04247. arXiv:2604.04247 [cs]

    work page internal anchor Pith review Pith/arXiv arXiv 2026

  23. [23]

    AgentGit: A version control framework for reliable and scalable LLM-powered multi-agent systems.arXiv preprint arXiv:2511.00628, 2025

    Yang Li, Siqi Ping, Xiyu Chen, Xiaojian Qi, Zigan Wang, Ye Luo, and Xiaowei Zhang. AgentGit: A Version Control Framework for Reliable and Scalable LLM-Powered Multi-Agent Systems, November 2025. URLhttps://arxiv.org/abs/2511.00628. arXiv:2511.00628

    work page arXiv 2025

  24. [24]

    Stop Wasting Your Tokens: Towards Efficient Runtime Multi-Agent Systems,

    Fulin Lin, Shaowen Chen, Ruishan Fang, Hongwei Wang, and Tao Lin. Stop wasting your tokens: Towards efficient runtime multi-agent systems.arXiv preprint arXiv:2510.26585, 2025. URLhttps://arxiv.org/abs/2510.26585

    work page arXiv 2025

  25. [25]

    Terminal-Bench: Benchmarking Agents on Hard, Realistic Tasks in Command Line Interfaces

    Mike A. Merrill, Alexander G. Shaw, Nicholas Carlini, Boxuan Li, Harsh Raj, Ivan Bercovich, Lin Shi, Jeong Yeon Shin, Thomas Walshe, E. Kelly Buchanan, Junhong Shen, Guanghao Ye, Haowei Lin, Jason Poulos, Maoyu Wang, Marianna Nezhurina, Jenia Jitsev, Di Lu, Orfeas Menis Mastromichalakis, Zhiwei Xu, Zizhao Chen, Yue Liu, Robert Zhang, Leon Liangyu Chen, An...

    work page internal anchor Pith review Pith/arXiv arXiv 2026

  26. [26]

    Alexander Novikov, Ngân V˜u, Marvin Eisenberger, Emilien Dupont, Po-Sen Huang, Adam Zsolt Wagner, Sergey Shirobokov, Borislav Kozlovskii, Francisco J. R. Ruiz, Abbas Mehrabian, M. Pawan Kumar, Abigail See, Swarat Chaudhuri, George Holland, Alex Davies, Sebastian Nowozin, Pushmeet Kohli, and Matej Balog. AlphaEvolve: A coding agent for scientific and algor...

    work page internal anchor Pith review Pith/arXiv arXiv 2025

  27. [27]

    NVIDIA Nemotron 3: Efficient and Open Intelligence

    NVIDIA. Nvidia nemotron 3: Efficient and open intelligence, 2025. URL https://arxiv. org/abs/2512.20856. White Paper

    work page internal anchor Pith review Pith/arXiv arXiv 2025

  28. [28]

    Generalizing verifiable instruction following,

    Valentina Pyatkin, Saumya Malik, Victoria Graf, Hamish Ivison, Shengyi Huang, Pradeep Dasigi, Nathan Lambert, and Hannaneh Hajishirzi. Generalizing verifiable instruction following,

    work page

  29. [29]

    URLhttps://arxiv.org/abs/2507.02833

    work page internal anchor Pith review Pith/arXiv arXiv

  30. [30]

    Qwen3.5: Towards native multimodal agents, February 2026

    Qwen Team. Qwen3.5: Towards native multimodal agents, February 2026. URL https: //qwen.ai/blog?id=qwen3.5

    work page 2026

  31. [31]

    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, 2024. URL https://arxiv.org/abs/ 2402.03300

    work page internal anchor Pith review Pith/arXiv arXiv 2024

  32. [32]

    Reflexion: Language Agents with Verbal Reinforcement Learning

    Noah Shinn, Federico Cassano, Edward Berman, Ashwin Gopinath, Karthik Narasimhan, and Shunyu Yao. Reflexion: Language Agents with Verbal Reinforcement Learning, March 2023. URLhttps://arxiv.org/abs/2303.11366. arXiv:2303.11366

    work page internal anchor Pith review Pith/arXiv arXiv 2023

  33. [33]

    Reinforcement learning: An introduction 1st edition.Exp

    RS Sutton and AG Barto. Reinforcement learning: An introduction 1st edition.Exp. Psychol. Learn. Mem. Cogn, 30:1302–1321, 1998

    work page 1998

  34. [34]

    Hindsight credit assign- ment for long-horizon LLM agents.arXiv preprint arXiv:2603.08754,

    Hui-Ze Tan, Xiao-Wen Yang, Hao Chen, Jie-Jing Shao, Yi Wen, Yuteng Shen, Weihong Luo, Xiku Du, Lan-Zhe Guo, and Yu-Feng Li. Hindsight credit assignment for long-horizon llm agents, 2026. URLhttps://arxiv.org/abs/2603.08754

    work page arXiv 2026

  35. [35]

    Counterfactual explanations without open- ing the black box: Automated decisions and the gdpr.Harvard Journal of Law & Technology, 31(2):841–887, 2018

    Sandra Wachter, Brent Mittelstadt, and Chris Russell. Counterfactual explanations without open- ing the black box: Automated decisions and the gdpr.Harvard Journal of Law & Technology, 31(2):841–887, 2018

    work page 2018

  36. [36]

    Fork, Explore, Commit: OS Primitives for Agentic Exploration, February 2026

    Cong Wang and Yusheng Zheng. Fork, Explore, Commit: OS Primitives for Agentic Exploration, February 2026. URLhttps://arxiv.org/abs/2602.08199. arXiv:2602.08199

    work page arXiv 2026

  37. [37]

    AgentSPEX: An Agent SPecification and EXecution Language

    Pengcheng Wang, Jerry Huang, Jiarui Yao, Rui Pan, Peizhi Niu, Yaowenqi Liu, Ruida Wang, Renhao Lu, Yuwei Guo, and Tong Zhang. AgentSPEX: An Agent SPecification and EXecution Language, 2026. URLhttps://arxiv.org/abs/2604.13346

    work page internal anchor Pith review Pith/arXiv arXiv 2026

  38. [38]

    A practitioner’s guide to multi-turn agentic rein- forcement learning.arXiv preprint arXiv:2510.01132,

    Ruiyi Wang and Prithviraj Ammanabrolu. A practitioner’s guide to multi-turn agentic reinforce- ment learning, 2025. URLhttps://arxiv.org/abs/2510.01132

    work page arXiv 2025

  39. [39]

    The OpenHands Software Agent SDK: A Composable and Extensible Foundation for Production Agents

    Xingyao Wang, Jiayi Pan, Binyuan Hui, et al. OpenHands V1: Event-sourced state management for multi-agent coding systems.MLSys, 2026. URL https://arxiv.org/abs/2511.03690. arXiv:2511.03690

    work page internal anchor Pith review Pith/arXiv arXiv 2026

  40. [40]

    ThetaEvolve: Test-time Learning on Open Problems,

    Yiping Wang, Shao-Rong Su, Zhiyuan Zeng, Eva Xu, Liliang Ren, Xinyu Yang, Zeyi Huang, Xuehai He, Luyao Ma, Baolin Peng, Hao Cheng, Pengcheng He, Weizhu Chen, Shuohang Wang, Simon Shaolei Du, and Yelong Shen. ThetaEvolve: Test-time Learning on Open Problems,

    work page

  41. [41]

    URLhttps://arxiv.org/abs/2511.23473

    work page internal anchor Pith review Pith/arXiv arXiv

  42. [42]

    Evo-Memory: Benchmarking LLM Agent Test-time Learning with Self-Evolving Memory

    Tianxin Wei, Noveen Sachdeva, Benjamin Coleman, Zhankui He, Yuanchen Bei, Xuying Ning, Mengting Ai, Yunzhe Li, Jingrui He, Ed H. Chi, Chi Wang, Shuo Chen, Fernando Pereira, Wang- Cheng Kang, and Derek Zhiyuan Cheng. Evo-Memory: Benchmarking LLM Agent Test-time Learning with Self-Evolving Memory, 2025. URLhttps://arxiv.org/abs/2511.20857

    work page internal anchor Pith review Pith/arXiv arXiv 2025

  43. [43]

    Group-evolving agents: Open-ended self-improvement via experience sharing, 2026

    Zhaotian Weng, Antonis Antoniades, Deepak Nathani, Zhen Zhang, Xiao Pu, and Xin Eric Wang. Group-Evolving Agents: Open-Ended Self-Improvement via Experience Sharing, 2026. URLhttps://arxiv.org/abs/2602.04837. 13

    work page arXiv 2026

  44. [44]

    AutoGen: Enabling Next-Gen LLM Applications via Multi-Agent Conversation

    Qingyun Wu, Gagan Bansal, Jieyu Zhang, Yiran Wu, Beibin Li, Erkang Zhu, Li Jiang, Xiaoyun Zhang, Shaokun Zhang, Jiale Liu, Ahmed Hassan Awadallah, Ryen W White, Doug Burger, and Chi Wang. AutoGen: Enabling Next-Gen LLM Applications via Multi-Agent Conversation, August 2023. URLhttps://arxiv.org/abs/2308.08155. arXiv:2308.08155

    work page internal anchor Pith review Pith/arXiv arXiv 2023

  45. [45]

    Agent0: Unleashing Self-Evolving Agents from Zero Data via Tool-Integrated Reasoning,

    Peng Xia, Kaide Zeng, Jiaqi Liu, Can Qin, Fang Wu, Yiyang Zhou, Caiming Xiong, and Huaxiu Yao. Agent0: Unleashing Self-Evolving Agents from Zero Data via Tool-Integrated Reasoning,

    work page

  46. [46]

    URLhttps://arxiv.org/abs/2511.16043

    work page arXiv

  47. [47]

    Lillicrap, Kenji Kawaguchi, and Michael Shieh

    Yuxi Xie, Anirudh Goyal, Wenyue Zheng, Min-Yen Kan, Timothy P. Lillicrap, Kenji Kawaguchi, and Michael Shieh. Monte carlo tree search boosts reasoning via iterative preference learning,

    work page

  48. [48]

    URLhttps://arxiv.org/abs/2405.00451

    work page arXiv

  49. [49]

    Recursive Language Models

    Alex L. Zhang, Tim Kraska, and Omar Khattab. Recursive language models, 2025. URL https://arxiv.org/abs/2512.24601

    work page internal anchor Pith review Pith/arXiv arXiv 2025

  50. [50]

    Zhang, Zhening Li, and Omar Khattab

    Alex L. Zhang, Zhening Li, and Omar Khattab. The Mismanaged Geniuses Hypothesis, 2026. URLhttps://alexzhang13.github.io/blog/2026/mgh/. Blog post

    work page 2026

  51. [51]

    AgenTracer: Who Is Inducing Failure in the LLM Agentic Systems?

    Guibin Zhang, Junhao Wang, Junjie Chen, Wangchunshu Zhou, Kun Wang, and Shuicheng Yan. Agentracer: Who is inducing failure in the llm agentic systems?, 2025. URL https: //arxiv.org/abs/2509.03312

    work page Pith review arXiv 2025

  52. [52]

    Darwin Godel Machine: Open-Ended Evolution of Self-Improving Agents

    Jenny Zhang, Shengran Hu, Cong Lu, Robert Lange, and Jeff Clune. Darwin Gödel Machine: Open-Ended Evolution of Self-Improving Agents, 2025. URL https://arxiv.org/abs/ 2505.22954

    work page internal anchor Pith review Pith/arXiv arXiv 2025

  53. [53]

    Hyperagents, 2026

    Jenny Zhang, Bingchen Zhao, Wannan Yang, Jakob Foerster, Jeff Clune, Minqi Jiang, Sam Devlin, and Tatiana Shavrina. Hyperagents, March 2026. URL https://arxiv.org/abs/ 2603.19461. arXiv:2603.19461

    work page arXiv 2026

  54. [54]

    Verbalized sampling: How to mitigate mode collapse and unlock llm diversity

    Jiayi Zhang, Simon Yu, Derek Chong, Anthony Sicilia, Michael R. Tomz, Christopher D. Manning, and Weiyan Shi. Verbalized sampling: How to mitigate mode collapse and unlock LLM diversity. InarXiv preprint arXiv:2510.01171, 2025. URL https://arxiv.org/abs/ 2510.01171

    work page internal anchor Pith review arXiv 2025

  55. [55]

    Agentic Context Engineering: Evolving Contexts for Self-Improving Language Models

    Qizheng Zhang, Changran Hu, Shubhangi Upasani, Boyuan Ma, Fenglu Hong, Vamsidhar Kamanuru, Jay Rainton, Chen Wu, Mengmeng Ji, Hanchen Li, Urmish Thakker, James Zou, and Kunle Olukotun. Agentic Context Engineering: Evolving Contexts for Self-Improving Language Models, 2025. URLhttps://arxiv.org/abs/2510.04618

    work page internal anchor Pith review Pith/arXiv arXiv 2025

  56. [56]

    MemRL: Self-Evolving Agents via Runtime Reinforcement Learning on Episodic Memory,

    Shengtao Zhang, Jiaqian Wang, Ruiwen Zhou, Junwei Liao, Yuchen Feng, Zhuo Li, Yujie Zheng, Weinan Zhang, Ying Wen, Zhiyu Li, Feiyu Xiong, Yutao Qi, Bo Tang, and Muning Wen. MemRL: Self-Evolving Agents via Runtime Reinforcement Learning on Episodic Memory,

    work page

  57. [57]

    URLhttps://arxiv.org/abs/2601.03192

    work page internal anchor Pith review Pith/arXiv arXiv

  58. [58]

    Language Agent Tree Search Unifies Reasoning Acting and Planning in Language Models

    Andy Zhou, Kai Yan, Michal Shlapentokh-Rothman, Haohan Wang, and Yu-Xiong Wang. Language Agent Tree Search Unifies Reasoning Acting and Planning in Language Models, October 2023. URLhttps://arxiv.org/abs/2310.04406. arXiv:2310.04406

    work page internal anchor Pith review Pith/arXiv arXiv 2023

  59. [59]

    2–3% of one turn

    Maciej ´Swiechowski, Konrad Godlewski, Bartosz Sawicki, and Jacek Ma ´ndziuk. Monte carlo tree search: a review of recent modifications and applications.Artificial Intelligence Review, 56(3):2497–2562, 2022. ISSN 1573-7462. doi: 10.1007/s10462-022-10228-y. URL http://dx.doi.org/10.1007/s10462-022-10228-y. A Limitations and Future Works A.1 Limitations Pro...

    work page doi:10.1007/s10462-022-10228-y 2022

  60. [60]

    This should be the default for the vast majority of agents on most ticks

    “none”— everything is fine, let the agent keep working. This should be the default for the vast majority of agents on most ticks. Over-intervention destroys progress

    work page

  61. [61]

    oh, the supervisor is nudging me

    “steer”— CHEAPEST intervention. The agent’s conversation is kept intact; we only append a new user message with your guidance so the agent sees it as “oh, the supervisor is nudging me”. Full conversation history and tool call context are preserved, KV cache is reused. Use this when: • the agent is broadly on task but drifting or about to make a minor wron...

    work page

  62. [62]

    redirect

    “redirect”— EXPENSIVE. The agent’s current session is aborted and a fresh opencode session starts with your guidance as message 1. The agent loses ALL memory of what it has explored, read, tried, or learned — it starts from scratch (but the files it already edited are still there on disk). Use this when: • the agent is stuck in an obvious loop (same tool,...

    work page

  63. [63]

    none”. • “Stuck in a loop

    “revert”— EXPENSIVE and destructive. Same as redirect on the LLM side (new session, lost memory) PLUS the sandbox filesystem is rolled back to the pre-run checkpoint. All files the agent edited are discarded. Use this ONLY when: • the agent wrote files that corrupt the repo (overwrote core code with garbage, introduced unrelated changes, broke syntax) • t...

    work page

  64. [64]

    (turn 1, session 0 only)Read README.md for the search policy, mechanism axes, and loop hard rules; readORIENTATION.mdfor the run parameters

    work page

  65. [65]

    (turn 1, every session)Read brief.md for the session index, frontier candidate, remaining budget, and the exact pending-batch and journal-pending file names this session must produce

    work page

  66. [66]

    (working turns)Use the read-only inspection tools (§F.2.3) to inspect prior runs; pick a base reference (frontier,baseline,promoted, arun_NNN, acand-XXX, or a Meta-Git scope ref); construct sibling variants withstage_variant, or by writing files directly under variants/session_NNN/vXX/workflow/; attach targeted_examples = {improve, protect, invariant}to each

    work page

  67. [67]

    Whenfinish_session returns control to the host, the proposer’s LLM session terminates and no chat-history context survives across sessions

    (handoff turn)Write hypothesis_logs/session_NNN.md with ### Findings, ### Hypotheses,### Considered & Rejected, and### Selected Batch sections; drop a session-fragment into journal_pending/session_NNN.md; write the manifest to pending_batches/session_NNN.json; callfinish_session. Whenfinish_session returns control to the host, the proposer’s LLM session t...

    work page

  68. [68]

    It is the host’s compact projection of 6candidates, failures, prior logs, and the handoff contract

    Read ‘brief.md‘ first. It is the host’s compact projection of 6candidates, failures, prior logs, and the handoff contract

    work page

  69. [69]

    Choose a set of prior sources as the batch base: ‘frontier‘, 8‘baseline‘, ‘promoted‘, a run id, a candidate id, or a Meta-Git 9scope ref/name

    work page

  70. [70]

    Create sibling variants from that set of bases using 11‘stage_variant‘, or by passing full ‘files‘/‘workflow_dir‘ entries 12in a batch manifest

    work page

  71. [71]

    These targeted checks are the 15preflight before hidden aggregate dev scoring

    Every variant must include explicit targeted examples with improve, 14protect, and invariant intent. These targeted checks are the 15preflight before hidden aggregate dev scoring

    work page

  72. [72]

    Call ‘run_counterfactual_batch‘ or ‘submit_counterfactual_batch ‘

    work page

  73. [73]

    Inspect aggregate outcomes, write the required 18‘experiment_logs/eXXX.md‘ with ‘## Outcome‘ and ‘## Next‘, then 19call ‘finish_session ‘. 20 21Rules: fix failure classes, not literal train examples; preserve the 22Agentic task shape; use valid train ids from brief.md, 23candidate_catalog.json, or traces/metrics; avoid near-duplicate prompt 24tweaks; trea...

    work page

  74. [74]

    show_history; read one trace.md of an interesting prior 35run; identify a failure CLASS (>=2 examples)

    ANALYZE. show_history; read one trace.md of an interesting prior 35run; identify a failure CLASS (>=2 examples)

    work page

  75. [75]

    Required before every batch

    HYPOTHESIZE. Required before every batch. Write 37hypotheses/hNNN_*.md with Branch from / Claim / Proposed 38change / Expected outcome / Why this differs from 39previous attempts / Cache consequence sections

    work page

  76. [76]

    branch(from_ref=...) to reset workflow/

    BRANCH. branch(from_ref=...) to reset workflow/

    work page

  77. [77]

    EDIT + check_workflow

    work page

  78. [78]

    43Provide explicit targeted_examples for every variant

    Prefer run_counterfactual_batch with several staged siblings. 43Provide explicit targeted_examples for every variant

    work page

  79. [79]

    Required between runs

    OBSERVE. Required between runs. Write observations/oNNN.md

    work page

  80. [80]

    Flat dev is not a stop condition; 46switch to a structurally different move

    Repeat until budget exhausted. Flat dev is not a stop condition; 46switch to a structurally different move. F.2.3 Proposer tool surface The proposer’s tools fall into four groups: filesystem inspection and editing, ledger inspection, effect- level introspection, and host-mediated evaluation. Table 14 lists the live surface; full JSON-Schema specifications...

    work page

Showing first 80 references.