pith. machine review for the scientific record. sign in

arxiv: 2605.01482 · v2 · submitted 2026-05-02 · 💻 cs.AI

Recognition: 2 theorem links

· Lean Theorem

Grounding Multi-Hop Reasoning in Structural Causal Models via Group Relative Policy Optimization

Yunhan Bu , Quan Zhang , Huaping Zhang , Guotong Geng , Chunxiao Gao , Askar Hamdulla , Juan Wang , Qiuchi Li
show 4 more authors
Baohua Zhang Shuai Lei Yunbo Cao Zhunchen Luo
Authors on Pith no claims yet

Pith reviewed 2026-05-11 02:12 UTC · model grok-4.3

classification 💻 cs.AI
keywords multi-hop fact verificationstructural causal modelsgroup relative policy optimizationreinforcement learningchain of thoughthallucination mitigationlarge language modelsinterpretable reasoning
0
0 comments X

The pith

Grounding multi-hop fact verification in a structural causal model and optimizing it with group relative policy optimization yields more accurate and less hallucinated reasoning than standard chain-of-thought methods.

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

The paper seeks to fix hallucinations and broken logic chains that appear when large language models attempt multi-hop fact verification across scattered evidence. It does so by building an explicit structural causal model that treats the verification task as constructing causal links between pieces of evidence and the claim being checked. A reinforcement-learning procedure called group relative policy optimization then tunes the length and structure of the resulting reasoning chain. The authors first observe that accuracy follows an inverted U-shape with chain length, so the optimizer learns to stop at the sweet spot. Experiments on the HoVer and EX-FEVER datasets show the combined approach beats prior methods while producing chains that are easier to inspect.

Core claim

The SCM-GRPO framework represents multi-hop verification as the construction of a structural causal model that encodes dependencies among evidence items and the target claim; group relative policy optimization then adjusts the policy that generates reasoning steps so that the model trades off depth against conciseness, producing shorter, more reliable chains than unguided chain-of-thought prompting.

What carries the argument

Structural causal model that represents causal dependencies between evidence and claims, paired with group relative policy optimization that learns to select concise yet sufficient reasoning paths.

If this is right

  • The approach delivers higher accuracy than existing baselines on the HoVer and EX-FEVER benchmarks.
  • Reasoning chains become both shorter and more correct once the optimizer respects the observed inverted-U relationship between length and performance.
  • Explicit causal modeling supplies a human-readable trace of how each evidence item supports or contradicts the claim.
  • Hallucinations decrease because each reasoning step must be justified inside the causal graph rather than generated freely.
  • The same optimization can be reused on other multi-hop tasks that suffer from excessive or insufficient chain length.

Where Pith is reading between the lines

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

  • The same causal-plus-RL pattern could be tested on multi-step mathematical reasoning or scientific literature synthesis where evidence must be linked causally.
  • If the method generalizes, it suggests that future language-model systems may need explicit causal scaffolding rather than relying only on statistical next-token prediction.
  • Applying the framework to live news-verification pipelines would provide a measurable test of whether causal grounding reduces real-world misinformation spread.

Load-bearing premise

Explicitly building a structural causal model of the verification process will capture the true dependencies without adding new modeling errors that cancel out the gains over ordinary chain-of-thought prompting.

What would settle it

On a held-out multi-hop verification set, if the SCM-GRPO method produces chains that are longer, less accurate, or contain more factual errors than plain chain-of-thought, the central claim would be refuted.

Figures

Figures reproduced from arXiv: 2605.01482 by Askar Hamdulla, Baohua Zhang, Chunxiao Gao, Guotong Geng, Huaping Zhang, Juan Wang, Qiuchi Li, Quan Zhang, Shuai Lei, Yunbo Cao, Yunhan Bu, Zhunchen Luo.

Figure 1
Figure 1. Figure 1: Empirical analysis of the relationship between CoT length and model performance across varying hop counts. The black line indicates verification accuracy, revealing an inverted U-shaped correlation, particularly in complex scenarios (3-hop). The bars represent the distribution of structural variables, highlighting the dependency between structural complexity and reasoning reliability. We investigate this p… view at source ↗
Figure 2
Figure 2. Figure 2: The overall architecture of the proposed framework. The training pipeline consists of two distinct stages: (1) SFT, where the base model is aligned with the SCM-based reasoning paradigm using the structured dataset D𝑠𝑡𝑟𝑢𝑐𝑡; and (2) GRPO Reinforcement Learning Optimization, where the policy model is refined via group-wise sampling and a composite reward function (𝑅𝑐, 𝑅𝑠, 𝑅𝑙) to enhance reasoning robustness … view at source ↗
Figure 3
Figure 3. Figure 3: The pipeline of automated data construction. The pro￾cess transforms standard multi-hop queries from the seed dataset (D𝑠𝑒𝑒𝑑) into structured causal inference paths. By leveraging structured prompts, a generator LLM explicitly produces SCM components (Exogenous Variables U, Endogenous Variables V, and Structural Functions F), which are then assembled and filtered to create the high-quality structure datase… view at source ↗
Figure 4
Figure 4. Figure 4: Structural Complexity and Performance Overview. (a) RLHF significantly reduces endogenous variables and paths compared to SFT. (b) Despite structural pruning, RLHF maintains superior accuracy (70.35% vs 69.16%). (c) RLHF exhibits a stable, compact distribution, whereas SFT shows numerous high-complexity outliers. A.1. Impact of Optimization on Structural Complexity (Figs. 4a-4c) We first examine the aggreg… view at source ↗
Figure 4
Figure 4. Figure 4: Structural Complexity and Performance Overview. (a) RLHF significantly reduces endogenous variables and paths compared to SFT. (b) Despite structural pruning, RLHF maintains superior accuracy (70.35% vs 69.16%). (c) RLHF exhibits a stable, compact distribution, whereas SFT shows numerous high-complexity outliers. A.1. Impact of Optimization on Structural Complexity (Figs. 4a-4c) We first examine the aggreg… view at source ↗
Figure 5
Figure 5. Figure 5: Relationship Between Variables and Causal Paths. The sharp slope for SFT (𝑦 = 0.64𝑥) indicates complexity explosion, while the flat slope for RLHF (𝑦 = 0.12𝑥) suggests efficient evidence integration. A.2. Relationship Between Structural Expansion and Topology ( view at source ↗
Figure 5
Figure 5. Figure 5: Relationship Between Variables and Causal Paths. The sharp slope for SFT (𝑦 = 0.64𝑥) indicates complexity explosion, while the flat slope for RLHF (𝑦 = 0.12𝑥) suggests efficient evidence integration. A.2. Relationship Between Structural Expansion and Topology ( [PITH_FULL_IMAGE:figures/full_fig_p012_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: Statistical Validation. (e) Differences in structural components are statistically significant (𝑝 < 1𝑒 − 30). (f) SFT shows high internal correlation (0.85) between variables and paths, while RLHF (0.16) effectively decouples them. A.3. Statistical Significance and Internal Correlations (Figs. 6a-6b) Finally, we verify the statistical validity of these structural shifts. Figure 6a confirms that the observe… view at source ↗
Figure 6
Figure 6. Figure 6: Statistical Validation. (e) Differences in structural components are statistically significant (𝑝 < 1𝑒 − 30). (f) SFT shows high internal correlation (0.85) between variables and paths, while RLHF (0.16) effectively decouples them. A.3. Statistical Significance and Internal Correlations (Figs. 6a-6b) Finally, we verify the statistical validity of these structural shifts. Figure 6a confirms that the observe… view at source ↗
Figure 7
Figure 7. Figure 7: Structural Efficiency Metrics. (g) RLHF shifts the focus to exogenous evidence (88.7%). (h) RLHF improves path efficiency (0.29 paths/var) significantly compared to SFT (0.60 paths/var). 13 view at source ↗
Figure 7
Figure 7. Figure 7: Structural Efficiency Metrics. (g) RLHF shifts the focus to exogenous evidence (88.7%). (h) RLHF improves path efficiency (0.29 paths/var) significantly compared to SFT (0.60 paths/var). 14 [PITH_FULL_IMAGE:figures/full_fig_p014_7.png] view at source ↗
read the original abstract

Multi-Hop Fact Verification (MHFV) necessitates complex reasoning across disparate evidence, posing significant challenges for Large Language Models (LLMs) which often suffer from hallucinations and fractured logical chains. Existing methods, while improving transparency via Chain-of-Thought (CoT), lack explicit modeling of the causal dependencies between evidence and claims. In this work, we introduce a novel framework that grounds reasoning in a Structural Causal Model (SCM), treating verification as a constructive causal inference process. We empirically identify an "inverted U-shaped" correlation between reasoning chain length and accuracy, revealing that excessive structural complexity degrades performance. To address this, we propose a Rule-based Reinforcement Learning strategy using Group Relative Policy Optimization (GRPO). This approach dynamically optimizes the trade-off between structural depth and conciseness. Extensive experiments on HoVer and EX-FEVER demonstrate that our SCM-GRPO framework significantly outperforms state-of-the-art baselines, offering a reliable and interpretable solution for complex fact verification.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

3 major / 2 minor

Summary. The manuscript introduces SCM-GRPO, a framework that grounds multi-hop fact verification in Structural Causal Models (SCMs) by treating verification as constructive causal inference. It identifies an inverted-U correlation between reasoning chain length and accuracy, proposes rule-based Group Relative Policy Optimization (GRPO) to optimize the depth-conciseness trade-off, and reports significant outperformance over state-of-the-art baselines on the HoVer and EX-FEVER datasets.

Significance. If the results hold under rigorous validation, the work could advance interpretable LLM reasoning by explicitly modeling causal dependencies between evidence and claims, offering a potential path to reduce hallucinations in complex verification tasks. The GRPO strategy for dynamically balancing structural complexity and the empirical identification of the inverted-U curve would be notable contributions to causal grounding in AI if they are shown to stem from correct causal assumptions rather than optimization artifacts.

major comments (3)
  1. §3 (Method): The construction and specification of the SCM (variables, edges, and interventions for evidence-claim dependencies) is insufficiently detailed. Without explicit description or validation of the causal graph (e.g., whether hand-specified or learned), it is impossible to rule out misspecification risks such as spurious edges, which could produce systematically incorrect counterfactuals and attribute gains to GRPO fitting rather than causal grounding.
  2. §4 (Experiments): The reported outperformance on HoVer and EX-FEVER lacks error bars, statistical significance tests, multiple-run averages, dataset statistics, and ablation studies isolating the SCM component from GRPO and base LLM effects. This is load-bearing for the central claim that SCM grounding yields more accurate, less hallucinated chains than standard CoT.
  3. Abstract and §5 (Discussion): The asserted inverted-U correlation between chain length and accuracy is presented without supporting figures, data points, or analysis demonstrating that GRPO specifically mitigates SCM-induced errors rather than other factors; this weakens the motivation and interpretability claims.
minor comments (2)
  1. Ensure consistent expansion of acronyms (MHFV, SCM, GRPO, CoT) on first use in the main body and abstract.
  2. Figure captions and table legends should explicitly state the number of runs and any controls used for the reported metrics.

Simulated Author's Rebuttal

3 responses · 0 unresolved

We thank the referee for the constructive and detailed comments, which help clarify areas where the manuscript can be strengthened. We address each major comment below and commit to revisions that improve the rigor and clarity of the work.

read point-by-point responses

  1. Referee: §3 (Method): The construction and specification of the SCM (variables, edges, and interventions for evidence-claim dependencies) is insufficiently detailed. Without explicit description or validation of the causal graph (e.g., whether hand-specified or learned), it is impossible to rule out misspecification risks such as spurious edges, which could produce systematically incorrect counterfactuals and attribute gains to GRPO fitting rather than causal grounding.

    Authors: We agree that §3 would benefit from greater detail on the SCM. In the revised manuscript we will expand the section to explicitly list the variables (e.g., evidence nodes, claim nodes, and latent causal factors), the directed edges representing dependencies, and the specific interventions used for counterfactual reasoning. We will also clarify the construction process (hand-specified from task structure versus data-driven) and include a discussion of potential misspecification risks together with how the subsequent GRPO stage interacts with the graph. revision: yes

  2. Referee: §4 (Experiments): The reported outperformance on HoVer and EX-FEVER lacks error bars, statistical significance tests, multiple-run averages, dataset statistics, and ablation studies isolating the SCM component from GRPO and base LLM effects. This is load-bearing for the central claim that SCM grounding yields more accurate, less hallucinated chains than standard CoT.

    Authors: We acknowledge the need for stronger statistical reporting. The revised experimental section will report means and standard deviations across multiple independent runs, include statistical significance tests (e.g., paired t-tests against baselines), provide dataset statistics, and add ablation studies that isolate the SCM component, the GRPO objective, and the base LLM. These additions will directly support the claim that the observed gains derive from the SCM grounding rather than other factors. revision: yes

  3. Referee: Abstract and §5 (Discussion): The asserted inverted-U correlation between chain length and accuracy is presented without supporting figures, data points, or analysis demonstrating that GRPO specifically mitigates SCM-induced errors rather than other factors; this weakens the motivation and interpretability claims.

    Authors: We will add the requested supporting material to §5. The revision will include figures plotting accuracy against chain length, the underlying data points, and an analysis that compares SCM-GRPO against ablated variants (SCM without GRPO, GRPO without SCM) to show that the optimization specifically reduces errors attributable to excessive structural complexity in the causal graph. revision: yes

Circularity Check

0 steps flagged

No circularity in derivation chain

full rationale

The paper proposes SCM-GRPO as a new framework that models multi-hop verification inside a structural causal model and applies rule-based GRPO to balance chain length against an empirically observed inverted-U accuracy curve. No equations, definitions, or optimization steps in the provided abstract or description reduce a claimed prediction or result to a fitted input or self-citation by construction. Performance is evaluated on external benchmarks (HoVer, EX-FEVER) rather than on quantities defined from the same fitted parameters. The derivation therefore remains self-contained against external data and does not exhibit any of the enumerated circularity patterns.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

Only the abstract is available; no explicit free parameters, axioms, or invented entities are stated in sufficient detail to populate the ledger.

pith-pipeline@v0.9.0 · 5504 in / 1123 out tokens · 40608 ms · 2026-05-11T02:12:57.703359+00:00 · methodology

Review history (2 revisions) →

discussion (0)

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

Lean theorems connected to this paper

Citations machine-checked in the Pith Canon. Every link opens the source theorem in the public Lean library.

What do these tags mean?
matches
The paper's claim is directly supported by a theorem in the formal canon.
supports
The theorem supports part of the paper's argument, but the paper may add assumptions or extra steps.
extends
The paper goes beyond the formal theorem; the theorem is a base layer rather than the whole result.
uses
The paper appears to rely on the theorem as machinery.
contradicts
The paper's claim conflicts with a theorem or certificate in the canon.
unclear
Pith found a possible connection, but the passage is too broad, indirect, or ambiguous to say the theorem truly supports the claim.

Reference graph

Works this paper leans on

80 extracted references · 80 canonical work pages · 16 internal anchors

  1. [1]

    GPT-4 Technical Report

    GPT-4 Technical Report , author =. arXiv preprint arXiv:2303.08774 , year =

    work page internal anchor Pith review Pith/arXiv arXiv

  2. [2]

    DeepSeek-V3 Technical Report

    DeepSeek-V3 Technical Report , author =. arXiv preprint arXiv:2412.19437 , year =

    work page internal anchor Pith review Pith/arXiv arXiv

  3. [3]

    Training a Helpful and Harmless Assistant with Reinforcement Learning from Human Feedback

    Training a Helpful and Harmless Assistant with Reinforcement Learning from Human Feedback , author =. arXiv preprint arXiv:2204.05862 , year =

    work page internal anchor Pith review Pith/arXiv arXiv

  4. [4]

    A Multitask, Multilingual, Multimodal Evaluation of ChatGPT on Reasoning, Hallucination, and Interactivity , author =. Proceedings of the 13th International Joint Conference on Natural Language Processing and the 3rd Conference of the Asia-Pacific Chapter of the Association for Computational Linguistics (Volume 1: Long Papers) , pages =. 2023 , publisher ...

    work page doi:10.18653/v1/2023.ijcnlp-main.45 2023

  5. [5]

    Proceedings of the AAAI Conference on Artificial Intelligence , volume =

    Graph of Thoughts: Solving Elaborate Problems with Large Language Models , author =. Proceedings of the AAAI Conference on Artificial Intelligence , volume =. 2024 , doi =

    work page 2024

  6. [6]

    Transactions on Machine Learning Research , year =

    Program of Thoughts Prompting: Disentangling Computation from Reasoning for Numerical Reasoning Tasks , author =. Transactions on Machine Learning Research , year =

    work page

  7. [7]

    Chain-of-Verification Reduces Hallucination in Large Language Models

    Chain-of-Verification Reduces Hallucination in Large Language Models , author =. Findings of the Association for Computational Linguistics: ACL 2024 , pages =. 2024 , publisher =. doi:10.18653/v1/2024.findings-acl.212 , url =

    work page doi:10.18653/v1/2024.findings-acl.212 2024

  8. [8]

    Advances in Neural Information Processing Systems , volume =

    AlpacaFarm: A Simulation Framework for Methods that Learn from Human Feedback , author =. Advances in Neural Information Processing Systems , volume =

    work page

  9. [9]

    Transactions of the Association for Computational Linguistics , volume =

    Causal Inference in Natural Language Processing: Estimation, Prediction, Interpretation and Beyond , author =. Transactions of the Association for Computational Linguistics , volume =. 2022 , doi =

    work page 2022

  10. [10]

    Retrieval-Augmented Generation for Large Language Models: A Survey

    Retrieval-Augmented Generation for Large Language Models: A Survey , author =. arXiv preprint arXiv:2312.10997 , year =

    work page internal anchor Pith review Pith/arXiv arXiv

  11. [11]

    Advances in Neural Information Processing Systems , volume =

    Causal Abstractions of Neural Networks , author =. Advances in Neural Information Processing Systems , volume =. 2021 , url =

    work page 2021

  12. [12]

    Transactions of the Association for Computational Linguistics , volume =

    A Survey on Automated Fact-Checking , author =. Transactions of the Association for Computational Linguistics , volume =. 2022 , doi =

    work page 2022

  13. [13]

    Proceedings of the 2024 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies (Volume 6: Industry Track) , pages =

    AnnoLLM: Making Large Language Models to Be Better Crowdsourced Annotators , author =. Proceedings of the 2024 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies (Volume 6: Industry Track) , pages =. 2024 , publisher =. doi:10.18653/v1/2024.naacl-industry.15 , url =

    work page doi:10.18653/v1/2024.naacl-industry.15 2024

  14. [14]

    Faithful Chain-of-Thought Reasoning , author =. Proceedings of the 13th International Joint Conference on Natural Language Processing and the 3rd Conference of the Asia-Pacific Chapter of the Association for Computational Linguistics (Volume 1: Long Papers) , pages =. 2023 , publisher =. doi:10.18653/v1/2023.ijcnlp-main.20 , url =

    work page doi:10.18653/v1/2023.ijcnlp-main.20 2023

  15. [15]

    A Survey on Hallucination in Large Language Models: Principles, Taxonomy, Challenges, and Open Questions

    A Survey on Hallucination in Large Language Models: Principles, Taxonomy, Challenges, and Open Questions , author =. arXiv preprint arXiv:2311.05232 , year =

    work page internal anchor Pith review arXiv

  16. [16]

    Active retrieval augmented generation

    Active Retrieval Augmented Generation , author =. Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing , pages =. 2023 , publisher =. doi:10.18653/v1/2023.emnlp-main.495 , url =

    work page doi:10.18653/v1/2023.emnlp-main.495 2023

  17. [17]

    Advances in Neural Information Processing Systems , volume =

    Large Language Models are Zero-Shot Reasoners , author =. Advances in Neural Information Processing Systems , volume =

    work page

  18. [18]

    Advances in Neural Information Processing Systems , volume =

    Retrieval-Augmented Generation for Knowledge-Intensive NLP Tasks , author =. Advances in Neural Information Processing Systems , volume =

    work page

  19. [19]

    Dor Muhlgay, Ori Ram, Inbal Magar, Yoav Levine, Nir Ratner, Yonatan Belinkov, Omri Abend, Kevin Leyton-Brown, Amnon Shashua, and Yoav Shoham

    Halo: Estimation and Reduction of Hallucinations in Open-Source Weak Large Language Models , author =. arXiv preprint arXiv:2308.11764 , year =

    work page arXiv

  20. [20]

    Let's Verify Step by Step

    Let's Verify Step by Step , author =. arXiv preprint arXiv:2305.20050 , year =

    work page internal anchor Pith review Pith/arXiv arXiv

  21. [21]

    Evaluating the Logical Reasoning Ability of ChatGPT and GPT-4

    Evaluating the Logical Reasoning Ability of ChatGPT and GPT-4 , author =. arXiv preprint arXiv:2304.03439 , year =

    work page arXiv

  22. [22]

    Andrey Malinin and Mark Gales

    SelfCheckGPT: Zero-Resource Black-Box Hallucination Detection for Generative Large Language Models , author =. Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing , pages =. 2023 , publisher =. doi:10.18653/v1/2023.emnlp-main.557 , url =

    work page doi:10.18653/v1/2023.emnlp-main.557 2023

  23. [23]

    In: Bouamor, H., Pino, J., Bali, K

    FActScore: Fine-grained Atomic Evaluation of Factual Precision in Long Form Text Generation , author =. Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing , pages =. 2023 , publisher =. doi:10.18653/v1/2023.emnlp-main.741 , url =

    work page doi:10.18653/v1/2023.emnlp-main.741 2023

  24. [24]

    Advances in Neural Information Processing Systems , volume =

    Training Language Models to Follow Instructions with Human Feedback , author =. Advances in Neural Information Processing Systems , volume =

    work page

  25. [25]

    Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers) , pages =

    Fact-Checking Complex Claims with Program-Guided Reasoning , author =. Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers) , pages =. 2023 , publisher =. doi:10.18653/v1/2023.acl-long.386 , url =

    work page doi:10.18653/v1/2023.acl-long.386 2023

  26. [26]

    Qwen Technical Report

    Qwen Technical Report , author =. arXiv preprint arXiv:2309.16609 , year =

    work page internal anchor Pith review Pith/arXiv arXiv

  27. [27]

    Advances in Neural Information Processing Systems , volume =

    Direct Preference Optimization: Your Language Model is Secretly a Reward Model , author =. Advances in Neural Information Processing Systems , volume =

    work page

  28. [28]

    DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language Models

    DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language Models , author =. arXiv preprint arXiv:2402.03300 , year =

    work page internal anchor Pith review Pith/arXiv arXiv

  29. [29]

    Findings of the Association for Computational Linguistics: EMNLP 2021 , pages =

    Retrieval Augmentation Reduces Hallucination in Conversation , author =. Findings of the Association for Computational Linguistics: EMNLP 2021 , pages =. 2021 , publisher =. doi:10.18653/v1/2021.findings-emnlp.320 , url =

    work page doi:10.18653/v1/2021.findings-emnlp.320 2021

  30. [30]

    The Llama 3 Herd of Models

    The Llama 3 Herd of Models , author =. arXiv preprint arXiv:2407.21783 , year =

    work page internal anchor Pith review Pith/arXiv arXiv

  31. [31]

    Llama 2: Open Foundation and Fine-Tuned Chat Models

    Llama 2: Open Foundation and Fine-Tuned Chat Models , author =. arXiv preprint arXiv:2307.09288 , year =

    work page internal anchor Pith review Pith/arXiv arXiv

  32. [32]

    International Conference on Learning Representations , year =

    Self-Consistency Improves Chain of Thought Reasoning in Language Models , author =. International Conference on Learning Representations , year =

    work page

  33. [33]

    Advances in Neural Information Processing Systems , volume =

    Chain-of-Thought Prompting Elicits Reasoning in Large Language Models , author =. Advances in Neural Information Processing Systems , volume =

    work page

  34. [34]

    Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics (Volume 3: System Demonstrations) , pages =

    OpenICL: An Open-Source Framework for In-context Learning , author =. Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics (Volume 3: System Demonstrations) , pages =. 2023 , publisher =. doi:10.18653/v1/2023.acl-demo.47 , url =

    work page doi:10.18653/v1/2023.acl-demo.47 2023

  35. [35]

    The Rise and Potential of Large Language Model Based Agents: A Survey

    The Rise and Potential of Large Language Model Based Agents: A Survey , author =. arXiv preprint arXiv:2309.07864 , year =

    work page internal anchor Pith review arXiv

  36. [36]

    Corrective Retrieval Augmented Generation

    Corrective Retrieval Augmented Generation , author =. arXiv preprint arXiv:2401.15884 , year =

    work page internal anchor Pith review arXiv

  37. [37]

    International Conference on Learning Representations , year =

    ReAct: Synergizing Reasoning and Acting in Language Models , author =. International Conference on Learning Representations , year =

    work page

  38. [38]

    Advances in Neural Information Processing Systems , volume =

    Tree of Thoughts: Deliberate Problem Solving with Large Language Models , author =. Advances in Neural Information Processing Systems , volume =

    work page

  39. [39]

    International Conference on Learning Representations , year =

    Making Retrieval-Augmented Language Models Robust to Irrelevant Context , author =. International Conference on Learning Representations , year =

    work page

  40. [40]

    Siren's Song in the AI Ocean: A Survey on Hallucination in Large Language Models

    Siren's Song in the AI Ocean: A Survey on Hallucination in Large Language Models , author =. arXiv preprint arXiv:2309.01219 , year =

    work page internal anchor Pith review arXiv

  41. [41]

    Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers) , pages =

    Verify-and-Edit: A Knowledge-Enhanced Chain-of-Thought Framework , author =. Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers) , pages =. 2023 , publisher =. doi:10.18653/v1/2023.acl-long.320 , url =

    work page doi:10.18653/v1/2023.acl-long.320 2023

  42. [42]

    virtual samples

    MQuAKE: Assessing Knowledge Editing in Language Models via Multi-Hop Questions , author =. Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing , pages =. 2023 , publisher =. doi:10.18653/v1/2023.emnlp-main.971 , url =

    work page doi:10.18653/v1/2023.emnlp-main.971 2023

  43. [43]

    International Conference on Learning Representations , year =

    Least-to-Most Prompting Enables Complex Reasoning in Large Language Models , author =. International Conference on Learning Representations , year =

    work page

  44. [44]

    Advances in Neural Information Processing Systems Datasets and Benchmarks Track , year =

    FEVEROUS: Fact Extraction and VERification Over Unstructured and Structured Information , author =. Advances in Neural Information Processing Systems Datasets and Benchmarks Track , year =

    work page

  45. [45]

    International Conference on Learning Representations , year =

    Self-RAG: Learning to Retrieve, Generate, and Critique through Self-Reflection , author =. International Conference on Learning Representations , year =

    work page

  46. [46]

    RAM: Recover Any 3D Human Motion in-the-Wild

    RAM: Recover Any 3D Human Motion in-the-Wild , author =. arXiv preprint arXiv:2603.19929 , year =

    work page internal anchor Pith review Pith/arXiv arXiv

  47. [47]

    Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages =

    Human Motion Instruction Tuning , author =. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages =

    work page

  48. [48]

    Proceedings of the AAAI Conference on Artificial Intelligence , volume =

    Multiple Human Motion Understanding , author =. Proceedings of the AAAI Conference on Artificial Intelligence , volume =. 2026 , doi =

    work page 2026

  49. [49]

    Proceedings of the 33rd ACM International Conference on Multimedia , pages =

    MoCount: Motion-Based Repetitive Action Counting , author =. Proceedings of the 33rd ACM International Conference on Multimedia , pages =. 2025 , doi =

    work page 2025

  50. [50]

    Proceedings of the 33rd ACM International Conference on Multimedia , pages =

    Graph Canvas for Controllable 3D Scene Generation , author =. Proceedings of the 33rd ACM International Conference on Multimedia , pages =

    work page

  51. [51]

    Explaining context length scaling and bounds for language models.arXiv preprint arXiv:2502.01481, 2025

    Intrinsic Entropy of Context Length Scaling in LLMs , author =. arXiv preprint arXiv:2502.01481 , year =

    work page arXiv

  52. [52]

    Proceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision , pages =

    CPSeg: Finer-grained Image Semantic Segmentation via Chain-of-Thought Language Prompting , author =. Proceedings of the IEEE/CVF Winter Conference on Applications of Computer Vision , pages =. 2024 , doi =

    work page 2024

  53. [53]

    Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages =

    CountLLM: Towards Generalizable Repetitive Action Counting via Large Language Model , author =. Proceedings of the IEEE/CVF Conference on Computer Vision and Pattern Recognition , pages =. 2025 , doi =

    work page 2025

  54. [54]

    Findings of the Association for Computational Linguistics: ACL 2025 , pages =

    Bayesian Optimization for Controlled Image Editing via LLMs , author =. Findings of the Association for Computational Linguistics: ACL 2025 , pages =. 2025 , publisher =. doi:10.18653/v1/2025.findings-acl.523 , url =

    work page doi:10.18653/v1/2025.findings-acl.523 2025

  55. [55]

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

    The Role of Deductive and Inductive Reasoning in Large Language Models , author =. Proceedings of the 63rd Annual Meeting of the Association for Computational Linguistics (Volume 1: Long Papers) , pages =. 2025 , publisher =. doi:10.18653/v1/2025.acl-long.820 , url =

    work page doi:10.18653/v1/2025.acl-long.820 2025

  56. [56]

    Learning an efficient optimizer via hybrid-policy sub-trajectory balance.arXiv preprint arXiv:2511.00543, 2025a

    Learning an Efficient Optimizer via Hybrid-Policy Sub-Trajectory Balance , author =. arXiv preprint arXiv:2511.00543 , year =

    work page arXiv

  57. [57]

    Proceedings of the IEEE/CVF International Conference on Computer Vision , pages =

    Is Meta-Learning Out? Rethinking Unsupervised Few-Shot Classification with Limited Entropy , author =. Proceedings of the IEEE/CVF International Conference on Computer Vision , pages =

    work page

  58. [58]

    Advances in Neural Information Processing Systems , volume =

    Scaling Law for Time Series Forecasting , author =. Advances in Neural Information Processing Systems , volume =

    work page

  59. [59]

    Findings of the Association for Computational Linguistics: EMNLP 2025 , pages =

    PAMN: Multi-phase Correlation Modeling for Contrast-Enhanced 3D Medical Image Retrieval , author =. Findings of the Association for Computational Linguistics: EMNLP 2025 , pages =. 2025 , publisher =. doi:10.18653/v1/2025.findings-emnlp.184 , url =

    work page doi:10.18653/v1/2025.findings-emnlp.184 2025

  60. [60]

    SSRN Electronic Journal , year =

    Rare-Earth Exposure and Bottlenecks in AI Data Centers: Cost and Schedule Risk , author =. SSRN Electronic Journal , year =. doi:10.2139/ssrn.6232318 , url =

    work page doi:10.2139/ssrn.6232318

  61. [61]

    arXiv preprint arXiv:2511.21772 , year =

    A Unified Metric Architecture for AI Infrastructure: A Cross-Layer Taxonomy Integrating Performance, Efficiency, and Cost , author =. arXiv preprint arXiv:2511.21772 , year =. doi:10.48550/arXiv.2511.21772 , url =

    work page doi:10.48550/arxiv.2511.21772

  62. [62]

    arXiv preprint arXiv:2512.14197 , year =

    Location-Robust Cost-Preserving Blended Pricing for Multi-Campus AI Data Centers , author =. arXiv preprint arXiv:2512.14197 , year =. doi:10.48550/arXiv.2512.14197 , url =

    work page doi:10.48550/arxiv.2512.14197

  63. [63]

    Proceedings of the AAAI Conference on Artificial Intelligence , volume =

    ReTrack: Evidence-Driven Dual-Stream Directional Anchor Calibration Network for Composed Video Retrieval , author =. Proceedings of the AAAI Conference on Artificial Intelligence , volume =. 2026 , doi =

    work page 2026

  64. [64]

    Proceedings of the AAAI Conference on Artificial Intelligence , volume =

    ENCODER: Entity Mining and Modification Relation Binding for Composed Image Retrieval , author =. Proceedings of the AAAI Conference on Artificial Intelligence , volume =. 2025 , doi =

    work page 2025

  65. [65]

    Proceedings of the AAAI Conference on Artificial Intelligence , volume =

    HABIT: Chrono-Synergia Robust Progressive Learning Framework for Composed Image Retrieval , author =. Proceedings of the AAAI Conference on Artificial Intelligence , volume =. 2026 , doi =

    work page 2026

  66. [66]

    Proceedings of the 33rd ACM International Conference on Multimedia , pages =

    OFFSET: Segmentation-Based Focus Shift Revision for Composed Image Retrieval , author =. Proceedings of the 33rd ACM International Conference on Multimedia , pages =. 2025 , doi =

    work page 2025

  67. [67]

    Proceedings of the AAAI Conference on Artificial Intelligence , volume =

    INTENT: Invariance and Discrimination-aware Noise Mitigation for Robust Composed Image Retrieval , author =. Proceedings of the AAAI Conference on Artificial Intelligence , volume =. 2026 , doi =

    work page 2026

  68. [68]

    ACM Transactions on Multimedia Computing, Communications and Applications , year =

    REFINE: Composed Video Retrieval via Shared and Differential Semantics Enhancement , author =. ACM Transactions on Multimedia Computing, Communications and Applications , year =

    work page

  69. [69]

    arXiv preprint arXiv:2604.01617 (2026)

    STABLE: Efficient Hybrid Nearest Neighbor Search via Magnitude-Uniformity and Cardinality-Robustness , author =. arXiv preprint arXiv:2604.01617 , year =

    work page arXiv

  70. [70]

    IEEE Transactions on Dependable and Secure Computing , pages =

    ERASE: Bypassing Collaborative Detection of AI Counterfeit via Comprehensive Artifacts Elimination , author =. IEEE Transactions on Dependable and Secure Computing , pages =. 2026 , doi =

    work page 2026

  71. [71]

    Proceedings of the 33rd ACM International Conference on Multimedia , pages =

    HUD: Hierarchical Uncertainty-Aware Disambiguation Network for Composed Video Retrieval , author =. Proceedings of the 33rd ACM International Conference on Multimedia , pages =. 2025 , doi =

    work page 2025

  72. [72]

    ICASSP 2025 -- 2025 IEEE International Conference on Acoustics, Speech and Signal Processing , pages =

    PAIR: Complementarity-Guided Disentanglement for Composed Image Retrieval , author =. ICASSP 2025 -- 2025 IEEE International Conference on Acoustics, Speech and Signal Processing , pages =. 2025 , organization =

    work page 2025

  73. [73]

    ICASSP 2025 -- 2025 IEEE International Conference on Acoustics, Speech and Signal Processing , pages =

    MEDIAN: Adaptive Intermediate-Grained Aggregation Network for Composed Image Retrieval , author =. ICASSP 2025 -- 2025 IEEE International Conference on Acoustics, Speech and Signal Processing , pages =. 2025 , organization =

    work page 2025

  74. [74]

    Proceedings of the AAAI Conference on Artificial Intelligence , volume =

    UniHOI: Unified Human-Object Interaction Understanding via Unified Token Space , author =. Proceedings of the AAAI Conference on Artificial Intelligence , volume =. 2026 , doi =

    work page 2026

  75. [75]

    Engineering Applications of Artificial Intelligence , volume =

    InstrucRobo: Object-Centric Multi-Instruction Decoupling Model for Explainable Robotic Manipulation , author =. Engineering Applications of Artificial Intelligence , volume =. 2026 , issn =

    work page 2026

  76. [76]

    Neurocomputing , volume =

    UniBVR: Balancing Visual and Reasoning Abilities in Unified 3D Scene Understanding , author =. Neurocomputing , volume =. 2026 , doi =

    work page 2026

  77. [77]

    IEEE Transactions on Services Computing , year =

    Multi-Objective Unlearning in Recommender Systems via Preference Guided Pareto Exploration , author =. IEEE Transactions on Services Computing , year =

    work page

  78. [78]

    Advances in Neural Information Processing Systems , volume =

    UltraRE: Enhancing RecEraser for Recommendation Unlearning via Error Decomposition , author =. Advances in Neural Information Processing Systems , volume =

    work page

  79. [79]

    MathAgent: Adversarial Evolution of Constraint Graphs for Mathematical Reasoning Data Synthesis

    MathAgent: Adversarial Evolution of Constraint Graphs for Mathematical Reasoning Data Synthesis , author =. 2026 , archivePrefix =. 2604.11188 , primaryClass =

    work page internal anchor Pith review Pith/arXiv arXiv 2026

  80. [80]

    Dynamic Sampling that Adapts: Self-Aware Iterative Data Persistent Optimization for Mathematical Reasoning

    Dynamic Sampling that Adapts: Iterative DPO for Self-Aware Mathematical Reasoning , author =. 2025 , archivePrefix =. 2505.16176 , primaryClass =

    work page internal anchor Pith review Pith/arXiv arXiv 2025