Computation and Language · arXiv:2611.16816 v1 · submitted 12 Nov 2026 · 33 references
Outcome-Mass Alignment: Controlling Length-Dependent Trajectory Influence in Multi-Turn LLM Agent Reinforcement Learning
The paper says
“To keep long interaction contexts tractable, stepwise group-relative training for multi-turn LLM agents materializes each source trajectory into multiple response rows and valid tokens.”
more of the abstract
This expansion creates an outer-accounting problem distinct from within-trajectory credit assignment: row-based outcome statistics can count one episode outcome multiple times, while token-level reduction can assign unequal aggregate objective mass across trajectories. We term these effects outcome-statistics multiplicity and objective-mass imbalance. We introduce Outcome-Mass Alignment (OMA), a two-component method that makes both accounting choices explicit. Outcome Alignment computes outcome statistics over unique source trajectories, while Mass Alignment controls aggregate objective mass after within-trajectory aggregation of optimizer-native token objectives. Neither redesigns the base optimizer's credit assignment, importance ratios, or clipping. Across matched ALFWorld GRPO comparisons at two model scales, where failed trajectories typically expand much more than successful ones, OMA improves matched-horizon endpoint success by 7.0-9.4 percentage points. Ordinary WebShop exhibits weak or reversed expansion-outcome coupling and correspondingly mixed effects, whereas in a success-oriented WebShop setting, WebShop-SOL, OMA improves success nAUC from 0.399 to 0.511. Component ablations and GiGPO and GSPO comparisons further support component separability and cross-optimizer compatibility. These results show that trajectory expansion always defines an implicit weighting rule, but its practical effect depends on how expansion couples to outcomes and optimizer-native updates.
from the abstract
Pith says
Machine review · deepseek-v4-flash · 13 Nov 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. OMA identifies a genuine outer-accounting problem in stepwise multi-turn LLM agent RL training; the paper is well-executed and honest about its limits, but the single-seed main experiments and mixed WebShop results mean the central claim is well-supported only for ALFWorld. This line is wrong
3 major objections · 11 minor
How often objections from reviews like this one are right: not yet measured
Not checked: figures, code, what the 33 references say.
The paper says, at §4.2, Table 1, and §4.7
The review did not quote this passage.
A classifier read this sentence of the paper as acknowledging the concern (probability 0.94):
“Appendix A adds two-seed experiments for ALFWorld-3B and WebShop-1.5B; these sum- maries provide limited evidence of run-to-run varia- tion rather than population-level estimates.”
Pith says: Objection 1 of 3 · major
The paper's strongest claim is the ALFWorld improvement, but the main experiments use only a single seed (seed 0). The authors acknowledge this limitation in §Limitations and provide two-seed appendix results for ALFWorld-3B and WebShop-1.5B. However, the two-seed results for ALFWorld-3B show that OMA improves endpoint success over Base in both seeds, but the magnitude varies (87.50% vs 85.16%), and …
more of objection 1
… the nAUC advantage over MA is small and inconsistent across seeds (OMA nAUC 60.79% vs 58.41%; MA nAUC 57.45% vs 59.08%). The paper's central quantitative claim (7.0-9.4 pp improvement) is based on the single-seed main results. Given that the authors themselves note that these summaries provide 'limited evidence of run-to-run variation rather than population-level estimates', the main-text claim should be explicitly qualified as a single-seed observation, and the two-seed appendix should be referenced in the abstract and conclusion. This does not invalidate the claim but is a load-bearing issue for the strength of the quantitative evidence.
The paper says, at §3.3, Eq. (8), and §B
The review did not quote this passage.
A classifier read this sentence of the paper as acknowledging the concern (probability 0.87):
“Equation (8) in the main pa- per gives the fractional-mass implementation used when a trajectory is split across actor microbatches, but arbitrary independently normalized repartition- ing is outside the invariance claim.”
Pith says: Objection 2 of 3 · major
Mass Alignment's invariance to arbitrary repartitioning is explicitly limited to complete-group reductions. The fractional-mass implementation in Eq. (8) handles partial reductions, but the authors state that 'arbitrary independently normalized repartitioning is outside the invariance claim' (§B). This is a reasonable scope condition, but it means that in practice, when trajectories are split across microbatches with separate normalization, the exact 1/K outer mass guarantee may not hold.
more of objection 2
The paper should discuss how sensitive the method is to this practical issue, e.g., whether the fractional-mass approximation is empirically close to the ideal, or whether there are known failure modes.
The paper says, at §4.4, Table 4
The review did not quote this passage.
Pith says: Objection 3 of 3 · major
The frozen-batch gradient audit is a nice diagnostic, but it uses only five representative batches (one WebShop, four ALFWorld) plus two additional rows. The selection criteria for these batches are not specified. Are they random? Are they chosen to span a range of expansion ratios? The paper should clarify how these batches were selected and whether they are representative of typical training batches. Without this, the audit is illustrative but not a systematic characterization.
11 minor objections
The paper says, at Abstract
Pith says: minor objection 1 of 11
The abstract states 'improves matched-horizon endpoint success by 7.0-9.4 percentage points' without noting that this is from single-seed experiments. Consider adding a qualifier such as 'in single-seed main experiments' or referencing the appendix.
The paper says, at §1, Figure 1
Pith says: minor objection 2 of 11
Figure 1 is very effective. However, the caption could be more self-contained; it currently refers to 'this pair' without specifying that the numbers come from the example in the text. Consider adding the token counts directly in the caption.
show 9 more
The paper says, at §3.1, Eq. (5)
Pith says: minor objection 3 of 11
The covariance identity is clear, but the notation Ei[ni] for the mean of ni over trajectories could be confused with expectation over a distribution. Since the group size K is fixed, this is fine, but a brief note that Ei is the empirical average over the group would help readers unfamiliar with this notation.
The paper says, at §4.1, §C.2
Pith says: minor objection 4 of 11
The nAUC definition uses linear interpolation between validation checkpoints. Since validation is every 5 updates, this is reasonable. However, the paper should state whether the nAUC is computed over the full [0,150] range or only at observed points. Eq. (12) in §C.2 clarifies this, but it would be helpful to mention in §4.1 that nAUC uses trapezoidal integration with linear interpolation.
The paper says, at §4.3, Table 3
Pith says: minor objection 5 of 11
The component ablation table shows that OA alone underperforms Base on ALFWorld (endpoint 61.7% vs 74.2% at 3B). This is an interesting result that deserves a brief explanation in the main text. The paper notes that OA 'removes row-population sensitivity but leaves token-weighted mass', but it would be helpful to explain why this leads to worse performance, perhaps referencing the covariance identity or the gradient audit.
The paper says, at §4.5, Figure 6
Pith says: minor objection 6 of 11
The learning dynamics figure is dense and the interpretation is relegated to the supplement. Consider adding a brief summary of the key takeaway in the main text, e.g., 'MA reduces entropy and shortens episodes earliest, consistent with its early-stage advantage, while OMA maintains higher entropy and achieves stronger late-stage performance.'
The paper says, at §5, Related Work
Pith says: minor objection 7 of 11
The discussion of AsyncWebRL (Bai et al., 2026) is helpful for positioning. However, the paper could also mention other length-aware methods such as Dr. GRPO (Liu et al., 2025) and λ-GRPO (Wang et al., 2025a) in more detail, as they are cited but not discussed in the context of length weighting.
The paper says, at §6, Conclusion
Pith says: minor objection 8 of 11
The conclusion states 'results support a condition-dependent effect' and summarizes the findings well. It could be strengthened by explicitly stating the practical recommendation: when expansion–outcome coupling is strong, use OMA; when it is weak or reversed, OMA may not help and could hurt. This is implicit but not stated as a guideline.
The paper says, at Appendix A, Table 5
Pith says: minor objection 9 of 11
The two-seed results for ALFWorld-3B show that OA has lower nAUC than Base for both seeds (44.86% and 44.64% vs 54.04% and 47.08%). This reinforces the main-text finding that OA alone is harmful. However, the standard deviations for Base nAUC (4.92) and OA nAUC (0.16) are very different; the OA nAUC is almost identical across seeds. This might be a coincidence or could indicate a bug; the authors should comment on whether this is expected.
The paper says, at Appendix G, Table 9
Pith says: minor objection 10 of 11
The audit provenance is commendable. However, the 'Inv.' column reports '12/12' for all rows, meaning all 12 invariance checks passed. It would be helpful to list what the 12 checks are (e.g., row duplication, token inflation, both, for each of OA, MA, OMA).
The paper says, at General
Pith says: minor objection 11 of 11
The paper uses the term 'outer-accounting' which is clear in context. However, the distinction between 'outer-accounting' and 'within-trajectory credit assignment' could be made even sharper in the introduction by explicitly stating that OMA does not address the question of which actions within a trajectory were good or bad, only how much influence each trajectory has overall.
What would settle it
Re-run the ALFWorld-3B GRPO comparison with 5 independent seeds instead of 1 (main) or 2 (appendix), and report mean and standard deviation of endpoint success and nAUC. If the 7.0–9.4 percentage point improvement is robust across seeds, the central claim is strengthened; if variance is high, the condition-dependent account still holds but the effect size is less certain.
from the skeptic's reading
How the review went
4 readings by deepseek-v4-flash · 13 Nov 2026
- Reader holds up, with conditions
- Skeptic partly agrees
- Editor wrote the verdict above
- Referee major revision: 3 major objections, 11 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 paper identifies a real and previously overlooked problem: when you materialize multi-turn agent trajectories stepwise (one response row per turn), the expansion creates two distinct outer-accounting biases—outcome-statistics multiplicity and objective-mass imbalance. The formalization is clean, the proposed OMA correction is principled (two independent components that don't touch the base optimizer's credit assignment), and the experimental design is thoughtful: two environments, two model scales, component ablations, cross-optimizer tests, and a frozen-batch gradient audit. The authors are transparent about limitations—single-seed main experiments, two-seed appendix, mixed results on ordinary WebShop, and the condition-dependent nature of the benefit. This is a solid contribution to the subfield of LLM agent training.
The main soft spot is the reliance on a single seed for the primary experiments. The two-seed appendix provides only limited additional evidence, and the authors acknowledge this. For a method that could have non-trivial variance, this is a real weakness. The mixed/negative results on ordinary WebShop are honestly reported and consistent with the paper's condition-dependent account, but they do limit the generality of the claimed benefits. The WebShop-SOL contrast changes multiple variables (model scale, termination), so it's not a clean causal test—the authors note this. The weakest load-bearing assumption is that the base optimizer's within-trajectory credit assignment is correct; OMA explicitly doesn't redesign it, so the method's value is conditional on that being reasonable. The paper acknowledges this as a scope condition, not a flaw.
Who this is for: researchers working on multi-turn LLM agent RL training, especially those using stepwise materialization with GRPO or similar group-relative methods. The paper deserves a serious referee; the core idea is novel and the execution is careful. I'd recommend acceptance after the authors address the seed variability concern (more seeds) and ideally release code/data for reproducibility. The paper is not a desk reject.
The strongest objection
The paper identifies a genuine outer-accounting problem in stepwise multi-turn LLM agent RL training, proposes a clean two-component correction (OMA), and validates it across multiple environments, model scales, and optimizers. The authors are transparent about limitations: single-seed main experiments, two-seed appendix, mixed results on ordinary WebShop, and the condition-dependent nature of the benefit. The mathematical derivation of outcome-statistics multiplicity and objective-mass imbalance is sound, and the experimental design includes component ablations, frozen-batch gradient audits, and cross-optimizer compatibility tests that support the central claim. The weakest assumption identified by the reader—that the base optimizer's within-trajectory credit assignment is correct—is acknowledged by the authors as a deliberate design choice (OMA does not redesign it), and the paper's results are conditional on this assumption being reasonable. This is not a flaw in the argument but a stated scope condition. The paper does not claim universal improvement; it provides a condition-dependent account and tests it in regimes where expansion–outcome coupling varies. No internal inconsistency or unsupported leap was found. The concern about single-seed experiments is partially addressed by the two-seed appendix, and the authors explicitly note this limitation. The central claim is well-supported for the ALFWorld setting, and the broader applicability is appropriately qualified.
The reader's verdict is CONDITIONAL, with the weakest assumption being that the base optimizer's credit assignment is correct. I agree this is a scope condition, but it is not a load-bearing concern because the paper explicitly states OMA does not redesign within-trajectory credit assignment and the results are conditional on the base optimizer being reasonable. The paper does not claim OMA fixes fundamental flaws in GRPO/GiGPO/GSPO; it claims OMA corrects a separate outer-accounting problem. The reader's other concerns (single seed, no code, mixed WebShop results) are valid limitations but are acknowledged by the authors and do not undermine the central claim for ALFWorld. The paper's own limitations section (b173–b179) is thorough. I find no significant objection that would change the verdict. The headline reflects that the paper's argument is internally consistent, the math is sound, and the experimental evidence, while limited in seed count, supports the claimed effect in the primary setting. The concrete test I propose would address the seed-variability concern without requiring a change in verdict.
The referee's report
The paper makes a clear and well-motivated contribution by identifying and formalizing a previously overlooked outer-accounting issue in stepwise multi-turn agent RL. The mathematical derivation is sound, the experimental design is thorough (two environments, two model scales, component ablations, cross-optimizer tests, frozen-batch gradient audit), and the authors are transparent about limitations. The central claim is well-supported for ALFWorld, and the condition-dependent account is appropriately qualified. The method is simple, principled, and compatible with existing optimizers, making it practically useful.
Pith's reading of the paper (a machine's summary, not the authors')
When training language models as agents that interact with environments over multiple turns, a common approach is to expand each full trajectory (a sequence of actions and observations) into separate training rows, one per turn. This stepwise materialization keeps context lengths manageable but introduces an accounting problem: a long failed trajectory produces many more rows and tokens than a short successful one, so it dominates both the outcome statistics (the failed outcome gets counted many times) and the objective mass (the loss is weighted by token count). The paper formalizes these as 'outcome-statistics multiplicity' and 'objective-mass imbalance'. The proposed solution, Outcome-Mass Alignment (OMA), has two independent components. Outcome Alignment computes outcome statistics (like mean and standard deviation for advantage normalization) over unique source trajectories rather than over expanded rows, so each episode outcome is counted once. Mass Alignment averages the policy objective within each trajectory and then across trajectories, giving each source trajectory equal total weight regardless of its length. OMA does not change the underlying optimizer's credit assignment, importance ratios, or clipping. Experiments on ALFWorld (embodied household tasks) show that OMA improves success rates by 7-9 percentage points across two model sizes. On WebShop (web shopping), where the expansion-outcome coupling is weaker, the effects are mixed, but a success-oriented variant shows clear gains. The paper also demonstrates compatibility with other group-relative optimizers (GiGPO, GSPO).
Core claim. The central claim is that stepwise materialization of multi-turn LLM agent trajectories creates two distinct outer-accounting problems (outcome-statistics multiplicity and objective-mass imbalance) that can bias training toward longer (often failed) trajectories, and that the proposed Outcome-Mass Alignment (OMA) method corrects these biases, leading to improved performance when expansion couples to outcomes. The paper states: 'systematic differences in row and token expansion between high- and low-outcome trajectories can make implicit trajectory reweighting consequential, and that OMA mitigates this effect through explicit trajectory-level accounting' (sentence b110). The strongest specific claim is that OMA improves ALFWorld matched-horizon endpoint success by 7.0-9.4 percentage points across two model scales (abstract, sentence b6).
The premise it rests on. The weakest load-bearing assumption is that the base optimizer's native within-trajectory credit assignment, importance ratios, and clipping mechanisms are correct and should be preserved unchanged. OMA explicitly does not redesign these (Section 3, paragraph 1: 'Neither redesigns the base optimizer's credit assignment, importance ratios, or clipping'). If the underlying credit assignment (e.g., GRPO's group-relative advantage) is itself flawed or suboptimal, OMA's corrections to outer accounting may not yield improvements, or could even compound errors. The paper's results are conditional on the base optimizer being reasonable. A second critical assumption is that trajectory identities are reliable and that the episode outcome attached to rows of one source is identical (Section B, sentence b265). In practice, this requires careful implementation and may not hold in all training setups.
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
To keep long interaction contexts tractable, stepwise group-relative training for multi-turn LLM agents materializes each source trajectory into multiple response rows and valid tokens. This expansion creates an outer-accounting problem distinct from within-trajectory credit assignment: row-based outcome statistics can count one episode outcome multiple times, while token-level reduction can assign unequal aggregate objective mass across trajectories. We term these effects outcome-statistics multiplicity and objective-mass imbalance. We introduce Outcome-Mass Alignment (OMA), a two-component method that makes both accounting choices explicit. Outcome Alignment computes outcome statistics over unique source trajectories, while Mass Alignment controls aggregate objective mass after within-trajectory aggregation of optimizer-native token objectives. Neither redesigns the base optimizer's credit assignment, importance ratios, or clipping. Across matched ALFWorld GRPO comparisons at two model scales, where failed trajectories typically expand much more than successful ones, OMA improves matched-horizon endpoint success by 7.0-9.4 percentage points. Ordinary WebShop exhibits weak or reversed expansion-outcome coupling and correspondingly mixed effects, whereas in a success-oriented WebShop setting, WebShop-SOL, OMA improves success nAUC from 0.399 to 0.511. Component ablations and GiGPO and GSPO comparisons further support component separability and cross-optimizer compatibility. These results show that trajectory expansion always defines an implicit weighting rule, but its practical effect depends on how expansion couples to outcomes and optimizer-native updates.
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 · 2026-11-13
The authors' main claim, as the classifier picked it out 0.97
“These results show that trajectory expansion always defines an implicit weighting rule, but its practical effect depends on how expansion couples to outcomes and optimizer-native updates.”
How the main claim is stated: asserted 1.00
What the abstract itself states, as probabilities: claims first 0.07 · states limitation 0.81 · human participants 0.04 · controlled comparison 0.94 · randomized assignment 0.03 · states quantitative result 0.99 · compares with prior results 0.95 · mentions code or data availability 0.03
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 |
|---|---|---|---|---|
| To keep long interaction contexts tractable, stepwise group-relative training for multi-turn LLM agents materializes each source trajectory into multiple response rows and valid tokens. | background 0.96 | 0.26 | 0.03 | asserted 1.00 |
| This expansion creates an outer-accounting problem distinct from within-trajectory credit assignment: row-based outcome statistics can count one episode outcome multiple times, while token-level reduction can assign unequal aggregate objective mass across trajectories. | background 0.57 | 0.35 | 0.06 | asserted 1.00 |
| We term these effects outcome-statistics multiplicity and objective-mass imbalance. | background 0.51 | 0.56 | 0.03 | asserted 0.93 |
| We introduce Outcome-Mass Alignment (OMA), a two-component method that makes both accounting choices explicit. | contribution 0.98 | 0.89 | 0.07 | asserted 0.99 |
| Outcome Alignment computes outcome statistics over unique source trajectories, while Mass Alignment controls aggregate objective mass after within-trajectory aggregation of optimizer-native token objectives. | method 0.87 | 0.84 | 0.10 | asserted 1.00 |
| Neither redesigns the base optimizer's credit assignment, importance ratios, or clipping. | method 0.68 | 0.69 | 0.05 | asserted 1.00 |
| Across matched ALFWorld GRPO comparisons at two model scales, where failed trajectories typically expand much more than successful ones, OMA improves matched-horizon endpoint success by 7.0-9.4 percentage points. | result 1.00 | 0.96 | 0.33 | asserted 1.00 |
| Ordinary WebShop exhibits weak or reversed expansion-outcome coupling and correspondingly mixed effects, whereas in a success-oriented WebShop setting, WebShop-SOL, OMA improves success nAUC from 0.399 to 0.511. | result 1.00 | 0.94 | 0.12 | asserted 1.00 |
| Component ablations and GiGPO and GSPO comparisons further support component separability and cross-optimizer compatibility. | result 0.97 | 0.93 | 0.05 | asserted 0.95 |
| These results show that trajectory expansion always defines an implicit weighting rule, but its practical effect depends on how expansion couples to outcomes and optimizer-native updates. | interpretation 1.00 | 0.81 | 0.74 | asserted 1.00 |
What kind of paper the classifier reports this is: method or tool 1.00 · evidence from observation or experiment 0.84
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, 2026-11-13: 1 request, 7,649 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 (6)
Figures are the authors', reproduced from the paper.
References (33)
- [1] Mohit Shridhar and Xingdi Yuan and Marc-Alexandre Cote and Yonatan Bisk and Adam Trischler and Matthew Hausknecht , title =. International Conference on Learning Representations , year = on Pith
- [2] Shunyu Yao and Howard Chen and John Yang and Karthik Narasimhan , title =. Advances in Neural Information Processing Systems , volume = on Pith
- [3] 2024 , url =. 2412.15115 , archivePrefix = on PithPith's reviewarXiv
- [4] Zhihong Shao and Peiyi Wang and Qihao Zhu and Runxin Xu and Junxiao Song and Xiao Bi and Haowei Zhang and Mingchuan Zhang and Y. K. Li and Y. Wu and Daya Guo , title =. 2024 , url =. 2402.03300 , archivePrefix = on PithPith's reviewarXiv
- [5] Lang Feng and Zhenghai Xue and Tingcong Liu and Bo An , title =. 2025 , url =. 2505.10978 , archivePrefix = on PithPith's reviewarXiv
- [6] Qiying Yu and Zheng Zhang and Ruofei Zhu and Yufeng Yuan and Xiaochen Zuo and Yu Yue and Weinan Dai and Tiantian Fan and Gaohong Liu and Lingjun Liu and Xin Liu and Haibin Lin and Zhiqi Lin and Bole Ma and Guangming Sheng and Yuxuan Tong and Chi Zhang and Mofan Zhang and Wang ... on PithPith's reviewarXiv
- [7] Zichen Liu and Changyu Chen and Wenjun Li and Penghui Qi and Tianyu Pang and Chao Du and Wee Sun Lee and Min Lin , title =. 2025 , url =. 2503.20783 , archivePrefix = on PithPith's reviewarXiv
- [8] Chujie Zheng and Shixuan Liu and Mingze Li and Xiong-Hui Chen and Bowen Yu and Chang Gao and Kai Dang and Yuqiong Liu and Rui Men and An Yang and Jingren Zhou and Junyang Lin , title =. 2025 , url =. 2507.18071 , archivePrefix = on PithPith's reviewarXiv
- [9] Zhiyuan Zeng and Jiameng Huang and Zhangyue Yin and Jiashuo Liu and Ziniu Li and Bingrui Li and Yuhao Wu and Yining Zheng and Ge Zhang and Wenhao Huang and Xipeng Qiu , title =. 2026 , url =. 2605.04077 , archivePrefix = on PithPith's reviewarXiv
- [10] Zhiyuan He and Xufang Luo and Yike Zhang and Yuqing Yang and Lili Qiu , title =. 2025 , url =. 2509.07558 , archivePrefix = on PitharXiv
- [11] Shuo He and Lang Feng and Qi Wei and Xin Cheng and Lei Feng and Bo An , title =. 2026 , url =. 2602.22817 , archivePrefix = on PitharXiv
- [12] Jiazheng Li and Yawei Wang and Qiaojing Yan and Yijun Tian and Zhichao Xu and Huan Song and Panpan Xu and Lin Lee Cheong , title =. Findings of the Association for Computational Linguistics:. 2026 , doi = on Pith
- [13] Qiuyi Qi and Tian Liang and Mutian Bao and Jinjian Zhang and Dongnan Liu and Wei Zhou and Linjian Mo and Ming Kong and Jie Liu and Feng Zhang and Qiang Zhu , title =. Proceedings of the 64th Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers... on Pith
- [14] Daoyu Wang and Qingchuan Li and Mingyue Cheng and Jie Ouyang and Shuo Yu and Qi Liu and Enhong Chen , title =. 2026 , url =. 2604.18401 , archivePrefix = on PithPith's reviewarXiv
- [15] Zefang Zong and Dingwei Chen and Yang Li and Qi Yi and Bo Zhou and Chengming Li and Bo Qian and Peng Chen and Jie Jiang , title =. 2026 , url =. 2601.04767 , archivePrefix = on PithPith's reviewarXiv
- [16] Zihan Wang and Kangrui Wang and Qineng Wang and Pingyue Zhang and Linjie Li and Zhengyuan Yang and Xing Jin and Kefan Yu and Minh Nhat Nguyen and Licheng Liu and Eli Gottlieb and Yiping Lu and Kyunghyun Cho and Jiajun Wu and Li Fei-Fei and Lijuan Wang and Yejin Choi and Manlin... on PithPith's reviewarXiv
- [17] Xufang Luo and Yuge Zhang and Zhiyuan He and Zilong Wang and Siyun Zhao and Dongsheng Li and Luna K. Qiu and Yuqing Yang , title =. 2025 , url =. 2508.03680 , archivePrefix = on PithPith's reviewarXiv
- [18] Hao Bai and Rui Yang and Chenlu Ye and Spencer Whitehead and Aviral Kumar and Tong Zhang , title =. 2026 , url =. 2606.05597 , archivePrefix = on PithPith's reviewarXiv
- [19] John Schulman and Filip Wolski and Prafulla Dhariwal and Alec Radford and Oleg Klimov , title =. 2017 , url =. 1707.06347 , archivePrefix = on PithPith's reviewarXiv
- [20] Arash Ahmadian and Chris Cremer and Matthias Gall. Back to Basics: Revisiting. 2024 , url =. 2402.14740 , archivePrefix = on PithPith's reviewarXiv
show 13 more
- [21] Shunyu Yao and Jeffrey Zhao and Dian Yu and Nan Du and Izhak Shafran and Karthik Narasimhan and Yuan Cao , title =. 2023 , url =. 2210.03629 , archivePrefix = on PithPith's reviewarXiv
- [22] Marwa Abdulhai and Isadora White and Charlie Snell and Charles Sun and Joey Hong and Yuexiang Zhai and Kelvin Xu and Sergey Levine , title =. 2023 , url =. 2311.18232 , archivePrefix = on PithPith's reviewarXiv
- [23] Yifei Zhou and Andrea Zanette and Jiayi Pan and Sergey Levine and Aviral Kumar , title =. 2024 , url =. 2402.19446 , archivePrefix = on PithPith's reviewarXiv
- [24] Zehan Qi and Xiao Liu and Iat Long Iong and Hanyu Lai and Xueqiao Sun and Wenyi Zhao and Yu Yang and Xinyue Yang and Jiadai Sun and Shuntian Yao and Tianjie Zhang and Wei Xu and Jie Tang and Yuxiao Dong , title =. 2025 , url =. 2411.02337 , archivePrefix = on PithPith's reviewarXiv
- [25] Yifei Zhou and Song Jiang and Yuandong Tian and Jason Weston and Sergey Levine and Sainbayar Sukhbaatar and Xian Li , title =. 2025 , url =. 2503.15478 , archivePrefix = on PithPith's reviewarXiv
- [26] Ruiyi Wang and Prithviraj Ammanabrolu , title =. 2025 , url =. 2510.01132 , archivePrefix = on PitharXiv
- [27] Chenliang Li and Adel Elmahdy and Alex Boyd and Zhongruo Wang and Siliang Zeng and Alfredo Garcia and Parminder Bhatia and Taha Kass-Hout and Cao Xiao and Mingyi Hong , title =. 2026 , url =. 2511.20718 , archivePrefix = on PithPith's reviewarXiv
- [28] Yining Wang and Jinman Zhao and Chuangxin Zhao and Shuhao Guan and Gerald Penn and Shinan Liu , title =. 2025 , url =. 2510.06870 , archivePrefix = on PitharXiv
- [29] Hui-Ze Tan and Xiao-Wen Yang and Hao Chen and Jie-Jing Shao and Yi Wen and Yuteng Shen and Weihong Luo and Xiku Du and Lan-Zhe Guo and Yu-Feng Li , title =. 2026 , url =. 2603.08754 , archivePrefix = on PitharXiv
- [30] Jiangweizhi Peng and Yuanxin Liu and Ruida Zhou and Charles Fleming and Zhaoran Wang and Alfredo Garcia and Mingyi Hong , title =. 2026 , url =. 2602.16165 , archivePrefix = on PithPith's reviewarXiv
- [31] Quan Wei and Siliang Zeng and Chenliang Li and William Brown and Oana Frunza and Wei Deng and Anderson Schneider and Yuriy Nevmyvaka and Yang Katie Zhao and Alfredo Garcia and Mingyi Hong , title =. 2025 , url =. 2505.11821 , archivePrefix = on PithPith's reviewarXiv
- [32] Eduardo Pignatelli and Johan Ferret and Matthieu Geist and Thomas Mesnard and Hado van Hasselt and Olivier Pietquin and Laura Toni , title =. 2024 , url =. 2312.01072 , archivePrefix = on PithPith's reviewarXiv
- [33] Chenchen Zhang , title =. 2026 , url =. 2604.09459 , archivePrefix = on PithPith's reviewarXiv
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