Pith. sign in

REVIEW 4 major objections 4 minor 3 cited by

Iterative Self-Incentivization Empowers Large Language Models as Agentic Searchers

T0 review · 4 major / 4 minor · reviewed 2026-08-07 · deepseek-v4-flash

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

Pith's one-line read EXSEARCH claims that an LLM can teach itself to search by treating trajectories as latent variables in a Generalized EM loop, outperforming larger baselines on multi-hop QA.

desk verdict A plausible empirical recipe for LLM agentic search, but the convergence proof doesn't cover the stochastic algorithm and the code is missing. read the letter →

arxiv 2505.20128 v1 pith:4SOY5PNE submitted 2025-05-26 cs.CL

Zhengliang Shi , Lingyong Yan , Dawei Yin , Suzan Verberne , Maarten de Rijke , Zhaochun Ren This is my paper
classification cs.CL
keywords agenticsearchexpectation-maximizationself-improvingLLMiterativeretrievalmulti-hopquestionansweringretrieval-augmentedgenerationimportancesamplingtrajectories
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

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

The reading

EXSEARCH sets out to prove that a single language model can teach itself to search well by treating each search session as a latent reasoning trajectory. The model decides what to retrieve, retrieves it from an external index, and records the evidence; then, in a Generalized Expectation-Maximization loop, it samples such trajectories, weights them by how likely they are to lead to the correct answer, and re-trains itself on the weighted trajectories. The paper argues this self-incentivized loop converges and, on four knowledge-intensive benchmarks, lifts exact-match scores by as much as 7.8 points over the strongest iterative baseline. If right, it would mean that continual self-improvement for retrieval-augmented reasoning can be obtained with a fairly simple weighted maximum-likelihood objective rather than costly reinforcement learning.

What carries the argument

The load-bearing mechanism is the self-incentivized EM loop built on the importance weight $w(z)\propto p(y\mid x,z;\theta_t)$. In the E-step, EXSEARCH samples full trajectories from the current distribution $p(z\mid x;\theta_t)$ and assigns each trajectory a scalar weight equal to the model's own likelihood of the gold answer given that trajectory; in the M-step, it maximizes $\mathbb{E}_{z\sim p(z\mid x;\theta_t)}[w(z)\log p(z\mid x;\theta)+w(z)\log p(y\mid x,z;\theta)]$, splitting the signal into learning-to-search and learning-to-answer losses. This importance-sampled evidence lower bound makes the intractable posterior over trajectories tractable, and the paper's convergence claim follows from monotonicity and boundedness of the sequence $\{\log p(y\mid x;\theta_t)\}$.

What would settle it

Take a trained EXSEARCH checkpoint and swap the retrieval corpus for one in which gold answers appear in unrelated, misleading documents. If the self-weight still strongly favors trajectories reaching the answer and downstream accuracy drops relative to oracle-weighted training, the self-likelihood signal is miscalibrated and the loop's benefit depends on the training corpus being benign.

Watch

Extended reading notes

Core claim

On the paper's own terms, the central discovery is that the intractable problem of learning to search can be recast as a latent-variable estimation problem solved by Generalized Expectation-Maximization. The search trajectory $z=\{(x_i,d_i,e_i)\}$ is treated as a latent variable, the LLM's marginal log-likelihood of the correct answer $\log p(y\mid x;\theta)$ is bounded from below by an importance-sampled ELBO, and the importance weight $w(z)\propto p(y\mid x,z;\theta_t)$ is computed with the same model that generated the trajectory. Maximizing the re-weighted joint log-likelihood in the M-step gives a self-incentivized loop: the current model scores its own trajectories by how well they support the gold answer, then trains itself to produce more of those trajectories. The paper claims this loop monotonically improves $\log p(y\mid x;\theta)$ and converges to a finite limit, and reports consistent gains across benchmark tasks and backbone models.

Load-bearing premise

The load-bearing premise is that the model's own probability for the gold answer, given a sampled trajectory, is a trustworthy measure of how good that trajectory is; if that self-score is miscalibrated, the EM loop can reinforce poor search instead of correcting it.

Editorial extensions

If this is right

  • A 7B-8B LLM trained this way can outperform much larger closed-book reasoners and specialized RAG systems on multi-hop QA benchmarks.
  • Because the reward signal is the model's own answer likelihood rather than an external reward model, the pipeline can be reproduced with standard supervised fine-tuning machinery and a frozen retriever.
  • The method scales across model families and sizes, and adding a re-ranking action to the trajectory yields further gains, suggesting the action space can be extended modularly.
  • The convergence argument implies that iterating the EM loop will not degrade the training objective, and empirical checkpoints peak within two or three iterations.
  • Failure modes the authors identify (over-searching and under-searching in 7.5% and 3.5% of examined cases) point to stopping criteria as the main open problem.

Reading between the lines

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

  • A testable extension would swap the self-likelihood weight for an oracle or human-judged trajectory quality signal; if that other signal materially outperforms the self-weight, the loop's reliance on self-calibration is the bottleneck.
  • The same EM formulation may transfer to tool-augmented agents, where tool calls are trajectory actions, but the current weight signal only measures answer-likelihood, not tool-use efficiency, so it may need a cost or latency term.
  • If the model's likelihood is miscalibrated, the self-incentivized loop could reinforce short, plausible trajectories that happen to yield the gold answer without genuinely retrieving useful evidence; monitoring retrieval recall across iterations would reveal whether this occurs.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

4 major / 4 minor

Summary. The paper introduces EXSEARCH, a framework that trains an LLM to interleave thinking, search, and recording in a multi-step agentic retrieval process. Training alternates between an E-step that samples search trajectories with the current LLM and assigns importance weights w(z) ∝ p(y|x,z;θ_t), and an M-step that maximizes a reweighted log-likelihood objective. The authors claim this is a Generalized Expectation-Maximization procedure with convergence guarantees, report large F1/EM/Acc gains over strong baselines on NQ, HotpotQA, MuSiQue, and 2WikiQA, and provide EXSEARCH-Zoo with additional backbones and a re-ranking action.

Significance. If the empirical claims hold, EXSEARCH is a useful step: it connects variational EM/self-training to agentic search, demonstrates gains across six backbone models, and carefully ablates thinking/search/recording and the importance weight. The paper ships extensive experiments, human evaluation, and an extension suite. However, the theoretical convergence guarantee is proven only for an idealized exact-EM procedure, not the stochastic dataset-level algorithm actually run, and the empirical convergence plots show evaluation-set degradation after the peak. These issues, together with the absence of variance estimates and a placeholder code link, mean the current version overstates both the theory and the reproducibility of the results.

major comments (4)
  1. [§3, Lemma 3.1 and Eq. (8)] The proof assumes θ_{t+1} is the exact per-example maximizer of ELBO(·, θ_t) and that the expectation in Eq. (5) is evaluated exactly. Algorithm 1 instead defines θ_{t+1} by maximizing the dataset-averaged objective E_{(x,y)~D}[L_R+L_A] and implements the M-step with stochastic gradient descent on a small number of sampled trajectories (two per example, Appendix D.5). Maximizing a sum of per-example ELBOs does not imply per-example monotonicity, and finite-sample SGD does not guarantee dataset-level monotonicity. Consequently, the chain in Eq. (8) does not prove convergence of the implemented EXSEARCH; it proves convergence only for an idealized exact-EM variant. Because contribution (ii) is stated in the abstract and Section 1 as a guarantee for the training process, this is a load-bearing gap. Please either provide a proof for the stochastic, dataset-level procedure under explicit assumptions or clearly restrict the theorem to the idealized variant.
  2. [Appendix C.3, Proposition 1] The proof of the claimed bound |arg max_θ ELBO(θ,q) − arg max_θ J(θ)| ≤ c·(KL(q ∥ r·p_θ))^{1/2} has several invalid steps: the KL term is written with an unnormalized target r·p_θ, which is not a probability distribution; L-smoothness of log J is assumed without argument; and the equality condition in Lemma C.5 ignores the normalization constant. The subsequent statement that w(z) ∝ r(y) makes the vanilla and goal-oriented objectives 'theoretically equivalent' is therefore not supported. Since this appendix is used to justify why the answer-likelihood weight is a principled training signal, the claim should be either proved carefully or removed.
  3. [§5.3 and Figure 3] The text says training 'consistently improves' and stabilizes, but the evaluation curves in Figure 3 peak in the second or third iteration and then decline for several dataset-model pairs (e.g., Qwen-2.5-7B and Llama-3.1-8B on HotpotQA and MuSiQue), and Algorithm 1 stops early based on validation performance. This is not the non-decreasing behavior predicted by Lemma 3.1 and weakens the empirical support for the convergence claim. Please report the curves with error bars over multiple runs and discuss the discrepancy explicitly.
  4. [Tables 1–3 and Appendix D.5] All main results are reported without variance or significance tests, and the claim that trajectory sampling numbers {1,2,4,6,8} make 'no significant difference' is not backed by a statistical test. The code link in Section 1 is a placeholder ('/gtbEXSEARCH'), so the experiments are not reproducible as described. For an empirical paper whose headline is a +7.8 EM improvement, please add variance/error bars and provide a usable code release.
minor comments (4)
  1. [§2.2 after Eq. (4)] The entropy constant is written as H(p(z|x;θt)) but should be H(p(z|x,y;θt)); as written it does not match Eq. (4).
  2. [§5.5 and Figure 4] '1,00 examples' appears twice and should read '100 examples'; 'approximately 2$' should read '$2' to match Appendix D.4.
  3. [Appendix F.2, Table 8] '100 randomly sampled cases from five benchmarks' conflicts with the four benchmarks used elsewhere in the paper.
  4. [Appendix F.1] The prompt uses the token '<FINIAL>' where the surrounding text and the final answer use '<FINAL>'.

Circularity Check

0 steps flagged · score 1.0 of 10

No significant circularity: the EM objective is anchored to external gold answers and evaluated on held-out benchmarks.

full rationale

EXSEARCH's derivation chain is anchored to an external objective, the marginal log-likelihood log p(y|x;θ) of gold answers, with trajectories treated as latent variables (Eqs. 2-6). The importance weight w(z)∝p(y|x,z;θ_t) is the current model's own likelihood, but it is used to reweight externally labeled gold answers, and all headline results are measured against held-out benchmark labels, so no fitted parameter is relabeled as a prediction. The convergence lemma (Lemma 3.1, Eq. 8) is the standard EM monotonicity argument; it is not circular, though it is a rigor gap that the proof assumes exact expectations and an exact per-example arg-max while Algorithm 1 performs dataset-level SGD with two sampled trajectories per example. Self-citations ([47], [96]) are baseline and evaluation references and are not load-bearing. No equation defines its target in terms of itself, and no unique derivation is imported from the authors' prior work.

Assumptions & free parameters 3 free parameters · 3 assumptions · 0 invented entities

The method relies on standard EM derivations and does not introduce new physical or conceptual entities. The only hand-chosen quantities are standard hyperparameters, which the paper shows are not sensitive.

free parameters (3)
  • Number of EM iterations N = 5
    Chosen by hand; performance on evaluation sets typically peaks at iterations 2-3, so the exact value is not load-bearing.
  • Number of sampled trajectories per example = 2
    Chosen by hand; the authors report varying this from 1 to 8 did not significantly change results.
  • Warm-up data size K = 1000
    Chosen by hand; the paper shows even K=0 yields improvements, so this is not central to the claim.
assumptions (3)
  • standard math Jensen's inequality yields a valid evidence lower bound.
    Used in Eq. (3) to derive the ELBO.
  • domain assumption The proposal distribution q(z|x) can be approximated by the prior p(z|x;θ_t) with importance sampling weights.
    The E-step samples from p(z|x;θ_t) and weights by w(z) ∝ p(y|x,z;θ_t); this assumes the importance weights adequately correct the distribution mismatch.
  • domain assumption The model can be optimized by stochastic gradient descent on the weighted objective.
    The M-step uses SGD; the convergence proof assumes exact maximization of the ELBO, which is not guaranteed by a single gradient step.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Iterative Self-Incentivization Empowers Large Language Models as Agentic Searchers." pith.science (2026). https://pith.science/paper/4SOY5PNE

@misc{pith2026250520128,
  author       = {Pith},
  title        = {Pith review of: Iterative Self-Incentivization Empowers Large Language Models as Agentic Searchers},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/4SOY5PNE}},
  note         = {Machine review of arXiv:2505.20128}
}
read the original abstract

Large language models (LLMs) have been widely integrated into information retrieval to advance traditional techniques. However, effectively enabling LLMs to seek accurate knowledge in complex tasks remains a challenge due to the complexity of multi-hop queries as well as the irrelevant retrieved content. To address these limitations, we propose EXSEARCH, an agentic search framework, where the LLM learns to retrieve useful information as the reasoning unfolds through a self-incentivized process. At each step, the LLM decides what to retrieve (thinking), triggers an external retriever (search), and extracts fine-grained evidence (recording) to support next-step reasoning. To enable LLM with this capability, EXSEARCH adopts a Generalized Expectation-Maximization algorithm. In the E-step, the LLM generates multiple search trajectories and assigns an importance weight to each; the M-step trains the LLM on them with a re-weighted loss function. This creates a self-incentivized loop, where the LLM iteratively learns from its own generated data, progressively improving itself for search. We further theoretically analyze this training process, establishing convergence guarantees. Extensive experiments on four knowledge-intensive benchmarks show that EXSEARCH substantially outperforms baselines, e.g., +7.8% improvement on exact match score. Motivated by these promising results, we introduce EXSEARCH-Zoo, an extension that extends our method to broader scenarios, to facilitate future work.

Figures

Figures reproduced from arXiv: 2505.20128 by the authors.

Figure 1
Figure 1. Performance on HotpotQA dataset when applying our E [PITH_FULL_IMAGE:figures/full_fig_p002_1.png] view at source ↗
Figure 2
Figure 2. Overall framework of the Expectation-Maximization process in [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. Training convergence of Qwen-2.5-7B and Llama-3.1-8B, where we report the Exact Match score for checkpoints in each iteration. training iteration on four experimental benchmarks [PITH_FULL_IMAGE:figures/full_fig_p008_3.png] view at source ↗
Figures from the paper (8 more)
Figure 4
Figure 4. Figure 4: Performance for Qwen2.5-7B that is initially [PITH_FULL_IMAGE:figures/full_fig_p008_4.png]
Figure 5
Figure 5. Figure 5: Performance of EXSEARCH when ex￾tended with the document re-ranking action. Extended Retrieval Strategy. In EXSEARCH, the LLM explores a search trajectory by itera￾tively thinking, searching, and recording. How￾ever, to answer more complex queries, we may need to custo…
Figure 6
Figure 6. Figure 6: Comparing vanilla training process in our method with variants trained using Exact Match [PITH_FULL_IMAGE:figures/full_fig_p023_6.png]
Figure 7
Figure 7. Figure 7: Training convergence of Qwen-2.5-7B and Llama-3.1-8B, where we report the F1 score for checkpoints in each iteration. The 0th iteration indicates the initial warm-up training. loss = loss * weights.mean() # Weight the loss by the average of the weights return loss [PI…
Figure 8
Figure 8. Figure 8: F1 score for EXSEARCH-Qwen-2.5-7B that was initially empowered by different amounts of warm-up data. low-resource and cold-start scenarios. Based on the cost analysis in § D.4, synthesizing each example costs only $0.02 on average. Thus, empirically, in our experiment,…
Figure 9
Figure 9. Figure 9: Exact match score for EXSEARCH-Qwen-2.5-7B, which is initially empowered by varying amounts of warm-up data, during the iterative training process in EXSEARCH. 32 [PITH_FULL_IMAGE:figures/full_fig_p032_9.png]
Figure 10
Figure 10. Figure 10: F1 score for EXSEARCH-Qwen-2.5-7B, which is initially empowered by varying amounts of warm-up data, during the iterative training process in EXSEARCH. 33 [PITH_FULL_IMAGE:figures/full_fig_p033_10.png]
Figure 11
Figure 11. Figure 11: Performance of EXSEARCH when extended with an additional document re-ranking action, where we report the exact match score, F1 score and accuracy for a comprehensive comparison. After finalizing a full trajectory z, the LLM generates a final answer y based on all inte…

Discussion (0). Sign in to comment.

Forward citations

Cited by 3 Pith papers

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

  1. Fetch-then-Explore: Decoupling Selection from Extraction over a Persistent Workspace for Search Agents

    cs.AI 2026-08 conditional novelty 6.0 of 10

    Fetch-then-Explore, which stores fetched pages in a per-question workspace and extracts evidence on demand with grep/read, beats visit-and-read and browsing baselines on BrowseComp across three LLM backbones.

  2. DeepShop: A Benchmark for Deep Research Shopping Agents

    cs.IR 2025-06 conditional novelty 6.0 of 10

    DeepShop, a benchmark of 150 complex online shopping queries with fine-grained evaluation, shows that leading web agents and deep research systems achieve at most a 32% task success rate.

  3. SIGMA: Search-Augmented On-Demand Knowledge Integration for Agentic Mathematical Reasoning

    cs.AI 2025-10 reject novelty 3.0 of 10

    SIGMA uses four specialized retrieval agents plus a moderator to improve math reasoning, reporting up to 7.4 absolute MATH500 points over Search-o1 at 1.5B scale.

Reference graph

Works this paper leans on

101 extracted references · 33 canonical work pages · cited by 3 Pith papers

  1. [1]

    Information retrieval on the web

    Mei Kobayashi and Koichi Takeda. Information retrieval on the web. ACM computing surveys (CSUR), 2000

    2000

  2. [2]

    A survey on rag meeting llms: Towards retrieval-augmented large language models

    Wenqi Fan, Yujuan Ding, Liangbo Ning, Shijie Wang, Hengyun Li, Dawei Yin, Tat-Seng Chua, and Qing Li. A survey on rag meeting llms: Towards retrieval-augmented large language models. In Proceedings of the 30th ACM SIGKDD Conference on Knowledge Discovery and Data Mining, pages 6491–6501, 2024

    2024

  3. [3]

    Retrieval augmented fact verification by synthesizing contrastive arguments

    Zhenrui Yue, Huimin Zeng, Lanyu Shang, Yifan Liu, Yang Zhang, and Dong Wang. Retrieval augmented fact verification by synthesizing contrastive arguments. In Proceedings of the 62nd Annual Meeting of the Association for Computational Linguistics, 2024

    2024

  4. [4]

    A survey on hallucination in large language models: Principles, taxonomy, challenges, and open questions

    Lei Huang, Weijiang Yu, Weitao Ma, Weihong Zhong, Zhangyin Feng, Haotian Wang, Qiang- long Chen, Weihua Peng, Xiaocheng Feng, Bing Qin, et al. A survey on hallucination in large language models: Principles, taxonomy, challenges, and open questions. ACM Transactions on Information Systems, 2023

    2023

  5. [5]

    Enhancing retrieval-augmented large language models with iterative retrieval-generation synergy

    Zhihong Shao, Yeyun Gong, Yelong Shen, Minlie Huang, Nan Duan, and Weizhu Chen. Enhancing retrieval-augmented large language models with iterative retrieval-generation synergy. In Findings of the Association for Computational Linguistics: EMNLP 2023, 2023

    2023

  6. [6]

    Is ChatGPT good at search? investigating large language models as re-ranking agents

    Weiwei Sun, Lingyong Yan, Xinyu Ma, Shuaiqiang Wang, Pengjie Ren, Zhumin Chen, Dawei Yin, and Zhaochun Ren. Is ChatGPT good at search? investigating large language models as re-ranking agents. In Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing, 2023

    2023

  7. [7]

    Recomp: Improving retrieval-augmented lms with compression and selective augmentation

    Fangyuan Xu, Weijia Shi, and Eunsol Choi. Recomp: Improving retrieval-augmented lms with compression and selective augmentation. arXiv preprint arXiv:2310.04408, 2023

    arXiv 2023

  8. [8]

    Retrieval-augmented generation for large language models: A survey

    Yunfan Gao, Yun Xiong, Xinyu Gao, Kangxiang Jia, Jinliu Pan, Yuxi Bi, Yi Dai, Jiawei Sun, and Haofen Wang. Retrieval-augmented generation for large language models: A survey. arXiv preprint arXiv:2312.10997, 2023

    arXiv 2023

Show all 101 references
  1. [9]

    Richrag: Crafting rich responses for multi-faceted queries in retrieval-augmented generation

    Shuting Wang, Xin Yu, Mang Wang, Weipeng Chen, Yutao Zhu, and Zhicheng Dou. Richrag: Crafting rich responses for multi-faceted queries in retrieval-augmented generation. arXiv preprint arXiv:2406.12566, 2024

    2024 arXiv

  2. [10]

    Self-rag: Learn- ing to retrieve, generate, and critique through self-reflection

    Akari Asai, Zeqiu Wu, Yizhong Wang, Avirup Sil, and Hannaneh Hajishirzi. Self-rag: Learn- ing to retrieve, generate, and critique through self-reflection. In The Twelfth International Conference on Learning Representations, ICLR 2024, Vienna, Austria, May 7-11, 2024, 2024

    2024

  3. [11]

    Improving scientific document retrieval with concept coverage-based query set generation

    SeongKu Kang, Bowen Jin, Wonbin Kweon, Yu Zhang, Dongha Lee, Jiawei Han, and Hwanjo Yu. Improving scientific document retrieval with concept coverage-based query set generation. In Proceedings of the Eighteenth ACM International Conference on Web Search and Data Mining, 2025. 10

    2025

  4. [12]

    Learning to explore and select for coverage-conditioned retrieval-augmented generation

    Takyoung Kim, Kyungjae Lee, Young Rok Jang, Ji Yong Cho, Gangwoo Kim, Minseok Cho, and Moontae Lee. Learning to explore and select for coverage-conditioned retrieval-augmented generation. arXiv preprint arXiv:2407.01158, 2024

    2024 arXiv

  5. [13]

    Making retrieval-augmented language models robust to irrelevant context

    Ori Yoran, Tomer Wolfson, Ori Ram, and Jonathan Berant. Making retrieval-augmented language models robust to irrelevant context. arXiv preprint arXiv:2310.01558, 2023

    2023 arXiv

  6. [14]

    Corrective retrieval augmented generation

    Shi-Qi Yan, Jia-Chen Gu, Yun Zhu, and Zhen-Hua Ling. Corrective retrieval augmented generation. arXiv preprint arXiv:2401.15884, 2024

    2024 arXiv

  7. [15]

    Rankrag: Unifying context ranking with retrieval-augmented generation in llms

    Yue Yu, Wei Ping, Zihan Liu, Boxin Wang, Jiaxuan You, Chao Zhang, Mohammad Shoeybi, and Bryan Catanzaro. Rankrag: Unifying context ranking with retrieval-augmented generation in llms. arXiv preprint arXiv:2407.02485, 2024

    2024 arXiv

  8. [16]

    Query rewriting in retrieval- augmented large language models

    Xinbei Ma, Yeyun Gong, Pengcheng He, Hai Zhao, and Nan Duan. Query rewriting in retrieval- augmented large language models. In Proceedings of the 2023 Conference on Empirical Methods in Natural Language Processing, 2023

    2023

  9. [17]

    Adelie: Aligning large language models on information extraction

    Yunjia Qi, Hao Peng, Xiaozhi Wang, Bin Xu, Lei Hou, and Juanzi Li. Adelie: Aligning large language models on information extraction. arXiv preprint arXiv:2405.05008, 2024

    2024 arXiv

  10. [18]

    Rag-ddr: Optimizing retrieval-augmented generation using differen- tiable data rewards

    Xinze Li, Sen Mei, Zhenghao Liu, Yukun Yan, Shuo Wang, Shi Yu, Zheni Zeng, Hao Chen, Ge Yu, Zhiyuan Liu, et al. Rag-ddr: Optimizing retrieval-augmented generation using differen- tiable data rewards. arXiv preprint arXiv:2410.13509, 2024

    arXiv

  11. [19]

    Importance sampling: a review

    Surya T Tokdar and Robert E Kass. Importance sampling: a review. Wiley Interdisciplinary Reviews: Computational Statistics, 2, 2010

    2010

  12. [20]

    Looking for information: A survey of research on information seeking, needs, and behavior

    Donald O Case and Lisa M Given. Looking for information: A survey of research on information seeking, needs, and behavior. Emerald Group Publishing, 2016

    2016

  13. [21]

    Agentic information retrieval

    Weinan Zhang, Junwei Liao, Ning Li, Kounianhua Du, and Jianghao Lin. Agentic information retrieval. arXiv preprint arXiv:2410.09713, 2024

    2024 arXiv

  14. [22]

    Exploratory search: from finding to understanding

    Gary Marchionini. Exploratory search: from finding to understanding. Commun. ACM, 49,

  15. [23]

    The expectation-maximization algorithm

    Todd K Moon. The expectation-maximization algorithm. IEEE Signal processing magazine, 1996

    1996

  16. [24]

    Variational reasoning about user preferences for conversa- tional recommendation

    Zhaochun Ren, Zhi Tian, Dongdong Li, Pengjie Ren, Liu Yang, Xin Xin, Huasheng Liang, Maarten de Rijke, and Zhumin Chen. Variational reasoning about user preferences for conversa- tional recommendation. In proceedings of the 45th international ACM SIGIR conference on research a...

    2022

  17. [25]

    Variational reasoning over incomplete knowledge graphs for conversational recommendation

    Xiaoyu Zhang, Xin Xin, Dongdong Li, Wenxuan Liu, Pengjie Ren, Zhumin Chen, Jun Ma, and Zhaochun Ren. Variational reasoning over incomplete knowledge graphs for conversational recommendation. In Proceedings of the Sixteenth ACM International Conference on Web Search and Data Mi...

    2023

  18. [26]

    Advances in importance sampling

    Víctor Elvira and Luca Martino. Advances in importance sampling. arXiv preprint arXiv:2102.05407, 2021

    2021 arXiv

  19. [27]

    Axiomatisations of the average and a further generalisation of monotonic sequences

    John Bibby. Axiomatisations of the average and a further generalisation of monotonic sequences. Glasgow Mathematical Journal, 1974

    1974

  20. [28]

    Retrieval-augmented generation for knowledge-intensive nlp tasks

    Patrick Lewis, Ethan Perez, Aleksandara Piktus, Fabio Petroni, Vladimir Karpukhin, Naman Goyal, Heinrich Kuttler, Mike Lewis, Wen tau Yih, Tim Rocktäschel, et al. Retrieval-augmented generation for knowledge-intensive nlp tasks. In Proceedings of the 34th International Confer-...

    2020

  21. [29]

    Search-in-the-chain: Towards accurate, credible and traceable large language models for knowledge-intensive tasks

    Shicheng Xu, Liang Pang, Huawei Shen, Xueqi Cheng, and Tat-Seng Chua. Search-in-the-chain: Towards accurate, credible and traceable large language models for knowledge-intensive tasks. In WWW, 2024. 11

    2024

  22. [30]

    Stochastic rag: End-to-end retrieval-augmented generation through expected utility maximization

    Hamed Zamani and Michael Bendersky. Stochastic rag: End-to-end retrieval-augmented generation through expected utility maximization. In Proceedings of the 47th International ACM SIGIR Conference on Research and Development in Information Retrieval, 2024

    2024

  23. [31]

    Investigating the factual knowledge boundary of large language models with retrieval augmentation

    Ruiyang Ren, Yuhao Wang, Yingqi Qu, Wayne Xin Zhao, Jing Liu, Hao Tian, Hua Wu, Ji-Rong Wen, and Haifeng Wang. Investigating the factual knowledge boundary of large language models with retrieval augmentation. In ACL, 2023

    2023

  24. [32]

    Natural questions: A benchmark for question answering research

    Tom Kwiatkowski, Jennimaria Palomaki, Olivia Redfield, Michael Collins, Ankur Parikh, Chris Alberti, Danielle Epstein, Illia Polosukhin, Jacob Devlin, Kenton Lee, et al. Natural questions: A benchmark for question answering research. Transactions of the Association for Computa...

    2019

  25. [33]

    Cohen, Ruslan Salakhut- dinov, and Christopher D

    Zhilin Yang, Peng Qi, Saizheng Zhang, Yoshua Bengio, William W. Cohen, Ruslan Salakhut- dinov, and Christopher D. Manning. HotpotQA: A dataset for diverse, explainable multi-hop question answering. In Conference on Empirical Methods in Natural Language Processing (EMNLP), 2018

    2018

  26. [34]

    MuSiQue: Multihop questions via single-hop question composition

    Harsh Trivedi, Niranjan Balasubramanian, Tushar Khot, and Ashish Sabharwal. MuSiQue: Multihop questions via single-hop question composition. In Transactions of the Association for Computational Linguistics: TACL, 2022

    2022

  27. [35]

    Constructing a multi-hop QA dataset for comprehensive evaluation of reasoning steps

    Xanh Ho, Anh-Khoa Duong Nguyen, Saku Sugawara, and Akiko Aizawa. Constructing a multi-hop QA dataset for comprehensive evaluation of reasoning steps. In Proceedings of the 28th International Conference on Computational Linguistics, 2020

    2020

  28. [36]

    KILT: a benchmark for knowledge intensive language tasks

    Fabio Petroni, Aleksandra Piktus, Angela Fan, Patrick Lewis, Majid Yazdani, Nicola De Cao, James Thorne, Yacine Jernite, Vladimir Karpukhin, Jean Maillard, et al. KILT: a benchmark for knowledge intensive language tasks. In Proceedings of the 2021 Conference of the North Amer-...

    2021

  29. [37]

    Deepseek-r1: Incentivizing reasoning capability in llms via reinforcement learning

    Daya Guo, Dejian Yang, Haowei Zhang, Junxiao Song, Ruoyu Zhang, Runxin Xu, Qihao Zhu, Shirong Ma, Peiyi Wang, Xiao Bi, et al. Deepseek-r1: Incentivizing reasoning capability in llms via reinforcement learning. arXiv preprint arXiv:2501.12948, 2025

    2025 arXiv

  30. [39]

    Introducing ChatGPT, 2022

    OpenAI. Introducing ChatGPT, 2022

    2022

  31. [40]

    An Yang, Baosong Yang, Beichen Zhang, Binyuan Hui, Bo Zheng, Bowen Yu, Chengyuan Li, Dayiheng Liu, Fei Huang, Haoran Wei, et al. Qwen2. 5 technical report. arXiv preprint arXiv:2412.15115, 2024

    2024 arXiv

  32. [41]

    Qwq: Reflect deeply on the boundaries of the unknown, 2024

    Qwen Team. Qwq: Reflect deeply on the boundaries of the unknown, 2024

    2024

  33. [42]

    The llama 3 herd of models

    Abhimanyu Dubey, Abhinav Jauhri, Abhinav Pandey, Abhishek Kadian, Ahmad Al-Dahle, Aiesha Letman, Akhil Mathur, Alan Schelten, Amy Yang, Angela Fan, et al. The llama 3 herd of models. arXiv preprint arXiv:2407.21783, 2024

    2024 arXiv

  34. [43]

    Mistral 7b

    Albert Qiaochu Jiang, Alexandre Sablayrolles, Arthur Mensch, Chris Bamford, Devendra Singh Chaplot, Diego de Las Casas, Florian Bressand, Gianna Lengyel, Guillaume Lample, Lucile Saulnier, et al. Mistral 7b. ArXiv, 2023

    2023

  35. [44]

    Chatqa: Surpassing gpt-4 on conversational qa and rag

    Zihan Liu, Wei Ping, Rajarshi Roy, Peng Xu, Chankyu Lee, Mohammad Shoeybi, and Bryan Catanzaro. Chatqa: Surpassing gpt-4 on conversational qa and rag. arXiv preprint arXiv:2401.10225, 2024

    2024 arXiv

  36. [45]

    Direct preference optimization: Your language model is secretly a reward model

    Rafael Rafailov, Archit Sharma, Eric Mitchell, Christopher D Manning, Stefano Ermon, and Chelsea Finn. Direct preference optimization: Your language model is secretly a reward model. Advances in Neural Information Processing Systems, 2024

    2024

  37. [46]

    InstructRAG: Instructing retrieval-augmented generation via self-synthesized rationales

    Zhepei Wei, Wei-Lin Chen, and Yu Meng. InstructRAG: Instructing retrieval-augmented generation via self-synthesized rationales. In The Thirteenth International Conference on Learning Representations, 2025. 12

    2025

  38. [47]

    Generate-then-ground in retrieval-augmented generation for multi-hop question answering

    Zhengliang Shi, Weiwei Sun, Shen Gao, Pengjie Ren, Zhumin Chen, and Zhaochun Ren. Generate-then-ground in retrieval-augmented generation for multi-hop question answering. arXiv preprint arXiv:2406.14891, 2024

    2024 arXiv

  39. [48]

    Khattab, Arnav Singhvi, Paridhi Maheshwari, Zhiyuan Zhang, Keshav Santhanam, Sri Vardhamanan, Saiful Haq, Ashutosh Sharma, Thomas T

    O. Khattab, Arnav Singhvi, Paridhi Maheshwari, Zhiyuan Zhang, Keshav Santhanam, Sri Vardhamanan, Saiful Haq, Ashutosh Sharma, Thomas T. Joshi, Hanna Moazam, et al. Dspy: Compiling declarative language model calls into state-of-the-art pipelines. In International Conference on ...

    2024

  40. [49]

    Verify-and-edit: A knowledge-enhanced chain-of-thought framework

    Ruochen Zhao, Xingxuan Li, Shafiq Joty, Chengwei Qin, and Lidong Bing. Verify-and-edit: A knowledge-enhanced chain-of-thought framework. arXiv preprint arXiv:2305.03268, 2023

    2023 arXiv

  41. [50]

    Generator-retriever-generator approach for open- domain question answering

    Abdelrahman Abdallah and Adam Jatowt. Generator-retriever-generator approach for open- domain question answering. arXiv preprint arXiv:2307.11278, 2023

    2023 arXiv

  42. [51]

    Search-o1: Agentic search-enhanced large reasoning models

    Xiaoxi Li, Guanting Dong, Jiajie Jin, Yuyao Zhang, Yujia Zhou, Yutao Zhu, Peitian Zhang, and Zhicheng Dou. Search-o1: Agentic search-enhanced large reasoning models. arXiv preprint arXiv:2501.05366, 2025

    2025 arXiv

  43. [52]

    Search-r1: Training llms to reason and leverage search engines with reinforcement learning

    Bowen Jin, Hansi Zeng, Zhenrui Yue, Dong Wang, Hamed Zamani, and Jiawei Han. Search-r1: Training llms to reason and leverage search engines with reinforcement learning. arXiv preprint arXiv:2503.09516, 2025

    2025 arXiv

  44. [53]

    Proximal policy optimization algorithms

    John Schulman, Filip Wolski, Prafulla Dhariwal, Alec Radford, and Oleg Klimov. Proximal policy optimization algorithms. arXiv preprint arXiv:1707.06347, 2017

    2017 arXiv

  45. [54]

    Khattab, Jon Saad-Falcon, Christopher Potts, and Matei A

    Keshav Santhanam, O. Khattab, Jon Saad-Falcon, Christopher Potts, and Matei A. Zaharia. Colbertv2: Effective and efficient retrieval via lightweight late interaction. In North American Chapter of the Association for Computational Linguistics, 2021

    2021

  46. [55]

    R1-searcher: Incentivizing the search capability in llms via reinforcement learning

    Huatong Song, Jinhao Jiang, Yingqian Min, Jie Chen, Zhipeng Chen, Wayne Xin Zhao, Lei Fang, and Ji-Rong Wen. R1-searcher: Incentivizing the search capability in llms via reinforcement learning. arXiv preprint arXiv:2503.05592, 2025

    2025 arXiv

  47. [56]

    Atm: Adversarial tuning multi- agent system makes a robust retrieval-augmented generator

    Junda Zhu, Lingyong Yan, Haibo Shi, Dawei Yin, and Lei Sha. Atm: Adversarial tuning multi- agent system makes a robust retrieval-augmented generator. arXiv preprint arXiv:2405.18111, 2024

    2024 arXiv

  48. [57]

    Improving retrieval-augmented generation through multi-agent reinforcement learning

    Yiqun Chen, Lingyong Yan, Weiwei Sun, Xinyu Ma, Yi Zhang, Shuaiqiang Wang, Dawei Yin, Yiming Yang, and Jiaxin Mao. Improving retrieval-augmented generation through multi-agent reinforcement learning. arXiv preprint arXiv:2501.15228, 2025

    2025

  49. [58]

    Deepspeed: System optimizations enable training deep learning models with over 100 billion parameters

    Jeff Rasley, Samyam Rajbhandari, Olatunji Ruwase, and Yuxiong He. Deepspeed: System optimizations enable training deep learning models with over 100 billion parameters. In Proceedings of the 26th ACM SIGKDD International Conference on Knowledge Discovery & Data Mining, 2020

    2020

  50. [59]

    Mixtral of experts

    Albert Q Jiang, Alexandre Sablayrolles, Antoine Roux, Arthur Mensch, Blanche Savary, Chris Bamford, Devendra Singh Chaplot, Diego de las Casas, Emma Bou Hanna, Florian Bressand, et al. Mixtral of experts. arXiv preprint arXiv:2401.04088, 2024

    2024 arXiv

  51. [60]

    Interleaving retrieval with chain-of-thought reasoning for knowledge-intensive multi-step questions

    Harsh Trivedi, Niranjan Balasubramanian, Tushar Khot, and Ashish Sabharwal. Interleaving retrieval with chain-of-thought reasoning for knowledge-intensive multi-step questions. In Proceedings of the 61st Annual Meeting of the Association for Computational Linguistics (Volume 1...

    2023

  52. [61]

    Vladimir Karpukhin, Barlas Oguz, Sewon Min, Patrick S. H. Lewis, Ledell Wu, Sergey Edunov, Danqi Chen, and Wen-tau Yih. Dense passage retrieval for open-domain question answering. In Proceedings of the 2020 Conference on Empirical Methods in Natural Language Processing, EMNLP ...

    2020

  53. [62]

    Document ranking with a pretrained sequence-to-sequence model

    Rodrigo Nogueira, Zhiying Jiang, Ronak Pradeep, and Jimmy Lin. Document ranking with a pretrained sequence-to-sequence model. In Findings of the Association for Computational Linguistics: EMNLP 2020, 2020. 13

    2020

  54. [63]

    C-pack: Packaged resources to advance general chinese embedding, 2023

    Shitao Xiao, Zheng Liu, Peitian Zhang, and Niklas Muennighoff. C-pack: Packaged resources to advance general chinese embedding, 2023

    2023

  55. [64]

    Rankvicuna: Zero-shot listwise document reranking with open-source large language models

    Ronak Pradeep, Sahel Sharifymoghaddam, and Jimmy Lin. Rankvicuna: Zero-shot listwise document reranking with open-source large language models. arXiv preprint arXiv:2309.15088, 2023

    2023 arXiv

  56. [65]

    Physics of language models: Part 3.1, knowledge storage and extraction

    Zeyuan Allen-Zhu and Yuanzhi Li. Physics of language models: Part 3.1, knowledge storage and extraction. arXiv preprint arXiv:2309.14316, 2023

    2023 arXiv

  57. [66]

    Information retrieval: recent advances and beyond

    Kailash A Hambarde and Hugo Proenca. Information retrieval: recent advances and beyond. IEEE Access, 11:76581–76604, 2023

    2023

  58. [67]

    Unsupervised information refinement training of large language models for retrieval-augmented generation

    Shicheng Xu, Liang Pang, Mo Yu, Fandong Meng, Huawei Shen, Xueqi Cheng, and Jie Zhou. Unsupervised information refinement training of large language models for retrieval-augmented generation. arXiv preprint arXiv:2402.18150, 2024

    2024 arXiv

  59. [68]

    Large language models for generative information extraction: A survey

    Derong Xu, Wei Chen, Wenjun Peng, Chao Zhang, Tong Xu, Xiangyu Zhao, Xian Wu, Yefeng Zheng, Yang Wang, and Enhong Chen. Large language models for generative information extraction: A survey. Frontiers of Computer Science, 18(6):186357, 2024

    2024

  60. [69]

    Demystifying long chain-of-thought reasoning in llms

    Edward Yeo, Yuxuan Tong, Morry Niu, Graham Neubig, and Xiang Yue. Demystifying long chain-of-thought reasoning in llms. arXiv preprint arXiv:2502.03373, 2025

    2025 arXiv

  61. [70]

    Towards reasoning era: A survey of long chain-of-thought for reasoning large language models

    Qiguang Chen, Libo Qin, Jinhao Liu, Dengyun Peng, Jiannan Guan, Peng Wang, Mengkang Hu, Yuhang Zhou, Te Gao, and Wangxiang Che. Towards reasoning era: A survey of long chain-of-thought for reasoning large language models. arXiv preprint arXiv:2503.09567, 2025

    2025 arXiv

  62. [71]

    Xia, Quoc Le, and Denny Zhou

    Jason Wei, Xuezhi Wang, Dale Schuurmans, Maarten Bosma, Ed Huai hsin Chi, F. Xia, Quoc Le, and Denny Zhou. Chain of thought prompting elicits reasoning in large language models. ArXiv, 2022

    2022

  63. [72]

    Reasoning with language model prompting: A survey

    Shuofei Qiao, Yixin Ou, Ningyu Zhang, Xiang Chen, Yunzhi Yao, Shumin Deng, Chuanqi Tan, Fei Huang, and Huajun Chen. Reasoning with language model prompting: A survey. arXiv preprint arXiv:2212.09597, 2022

    2022 arXiv

  64. [73]

    Towards large reasoning models: A survey of reinforced reasoning with large language models

    Fengli Xu, Qianyue Hao, Zefang Zong, Jingwei Wang, Yunke Zhang, Jingyi Wang, Xiaochong Lan, Jiahui Gong, Tianjian Ouyang, Fanjin Meng, et al. Towards large reasoning models: A survey of reinforced reasoning with large language models. arXiv preprint arXiv:2501.09686, 2025

    2025 arXiv

  65. [74]

    Reasoning with large language models, a survey

    Aske Plaat, Annie Wong, Suzan Verberne, Joost Broekens, Niki van Stein, and Thomas Back. Reasoning with large language models, a survey. arXiv preprint arXiv:2407.11511, 2024

    2024

  66. [75]

    Thinking, fast and slow

    Daniel Kahneman. Thinking, fast and slow. Farrar, Straus and Giroux, New York, 2011

    2011

  67. [76]

    Learning more effective representations for dense retrieval through deliberate thinking before search

    Yifan Ji, Zhipeng Xu, Zhenghao Liu, Yukun Yan, Shi Yu, Yishan Li, Zhiyuan Liu, Yu Gu, Ge Yu, and Maosong Sun. Learning more effective representations for dense retrieval through deliberate thinking before search. arXiv preprint arXiv:2502.12974, 2025

    2025

  68. [77]

    Tree of thoughts: Deliberate problem solving with large language models

    Shunyu Yao, Dian Yu, Jeffrey Zhao, Izhak Shafran, Tom Griffiths, Yuan Cao, and Karthik Narasimhan. Tree of thoughts: Deliberate problem solving with large language models. Ad- vances in neural information processing systems, 36, 2023

    2023

  69. [78]

    Deepseekmath: Pushing the limits of mathematical reasoning in open language models

    Zhihong Shao, Peiyi Wang, Qihao Zhu, Runxin Xu, Junxiao Song, Xiao Bi, Haowei Zhang, Mingchuan Zhang, YK Li, Y Wu, et al. Deepseekmath: Pushing the limits of mathematical reasoning in open language models. arXiv preprint arXiv:2402.03300, 2024

    2024 arXiv

  70. [79]

    Program of thoughts prompting: Disentangling computation from reasoning for numerical reasoning tasks

    Wenhu Chen, Xueguang Ma, Xinyi Wang, and William W Cohen. Program of thoughts prompting: Disentangling computation from reasoning for numerical reasoning tasks. arXiv preprint arXiv:2211.12588, 2022

    2022 arXiv

  71. [80]

    Torl: Scaling tool-integrated rl

    Xuefeng Li, Haoyang Zou, and Pengfei Liu. Torl: Scaling tool-integrated rl. arXiv preprint arXiv:2503.23383, 2025. 14

    2025 arXiv

  72. [81]

    What makes large language models reason in (multi-turn) code generation? arXiv preprint arXiv:2410.08105, 2024

    Kunhao Zheng, Juliette Decugis, Jonas Gehring, Taco Cohen, Benjamin Negrevergne, and Gabriel Synnaeve. What makes large language models reason in (multi-turn) code generation? arXiv preprint arXiv:2410.08105, 2024

    2024 arXiv

  73. [82]

    A survey on large language models for code generation

    Juyong Jiang, Fan Wang, Jiasi Shen, Sungju Kim, and Sunghun Kim. A survey on large language models for code generation. arXiv preprint arXiv:2406.00515, 2024

    2024 arXiv

  74. [83]

    In-context retrieval-augmented language models

    Ori Ram, Yoav Levine, Itay Dalmedigos, Dor Muhlgay, Amnon Shashua, Kevin Leyton-Brown, and Yoav Shoham. In-context retrieval-augmented language models. Transactions of the Association for Computational Linguistics, 11, 2023

    2023

  75. [84]

    Multi-level information retrieval augmented generation for knowledge-based visual question answering

    Omar Adjali, Olivier Ferret, Sahar Ghannay, and Hervé Le Borgne. Multi-level information retrieval augmented generation for knowledge-based visual question answering. In Proceedings of the 2024 Conference on Empirical Methods in Natural Language Processing, 2024

    2024

  76. [85]

    Where did my optimum go?: An em- pirical analysis of gradient descent optimization in policy gradient methods

    Peter Henderson, Joshua Romoff, and Joelle Pineau. Where did my optimum go?: An em- pirical analysis of gradient descent optimization in policy gradient methods. arXiv preprint arXiv:1810.02525, 2018

    2018 arXiv

  77. [86]

    The 37 implementation details of proximal policy optimization

    Shengyi Huang, Rousslan Fernand Julien Dossa, Antonin Raffin, Anssi Kanervisto, and Weixun Wang. The 37 implementation details of proximal policy optimization. In The ICLR Blog Track 2023, 2022

    2023

  78. [87]

    Rethinking the implementation tricks and monotonicity constraint in cooperative multi-agent reinforcement learning

    Jian Hu, Siyang Jiang, Seth Austin Harding, Haibin Wu, and Shih-wei Liao. Rethinking the implementation tricks and monotonicity constraint in cooperative multi-agent reinforcement learning. arXiv preprint arXiv:2102.03479, 2021

    2021 arXiv

  79. [88]

    Language models are hidden reasoners: Unlocking latent reasoning capabilities via self-rewarding

    Haolin Chen, Yihao Feng, Zuxin Liu, Weiran Yao, Akshara Prabhakar, Shelby Heinecke, Ricky Ho, Phil Mui, Silvio Savarese, Caiming Xiong, et al. Language models are hidden reasoners: Unlocking latent reasoning capabilities via self-rewarding. arXiv preprint arXiv:2411.04282, 2024

    2024 arXiv

  80. [89]

    Meta-rewarding language models: Self-improving alignment with llm-as-a-meta-judge

    Tianhao Wu, Weizhe Yuan, Olga Golovneva, Jing Xu, Yuandong Tian, Jiantao Jiao, Jason Weston, and Sainbayar Sukhbaatar. Meta-rewarding language models: Self-improving alignment with llm-as-a-meta-judge. arXiv preprint arXiv:2407.19594, 2024

    2024 arXiv

  81. [90]

    Co-Reyes, Rishabh Agarwal, Ankesh Anand, Piyush Patil, Peter J

    Avi Singh, John D. Co-Reyes, Rishabh Agarwal, Ankesh Anand, Piyush Patil, Peter J. Liu, James Harrison, Jaehoon Lee, Kelvin Xu, Aaron T Parisi, et al. Beyond human data: Scaling self-training for problem-solving with language models. Trans. Mach. Learn. Res., 2023

    2023

  82. [91]

    Buy 4 reinforce samples, get a baseline for free! In The International Conference on Learning Representations, 2019

    Wouter Kool, Herke van Hoof, and Max Welling. Buy 4 reinforce samples, get a baseline for free! In The International Conference on Learning Representations, 2019

    2019

  83. [92]

    Sample efficient reinforce- ment learning with reinforce

    Junzi Zhang, Jongho Kim, Brendan O’Donoghue, and Stephen Boyd. Sample efficient reinforce- ment learning with reinforce. In Proceedings of the AAAI Conference on Artificial Intelligence, 2021

    2021

  84. [93]

    Self-play fine-tuning converts weak language models to strong language models

    Zixiang Chen, Yihe Deng, Huizhuo Yuan, Kaixuan Ji, and Quanquan Gu. Self-play fine-tuning converts weak language models to strong language models. arXiv preprint arXiv:2401.01335, 2024

    2024 arXiv

  85. [94]

    Peter Dayan and Geoffrey E. Hinton. Using expectation-maximization for reinforcement learning. Neural Computation, 1997

    1997

  86. [95]

    Corpuslm: Towards a unified language model on corpus for knowledge-intensive tasks

    Xiaoxi Li, Zhicheng Dou, Yujia Zhou, and Fangchao Liu. Corpuslm: Towards a unified language model on corpus for knowledge-intensive tasks. In Proceedings of the 2024 ACM SIGIR International Conference on Theory of Information Retrieval, 2024

    2024

  87. [96]

    Rade: Reference-assisted dialogue evaluation for open-domain dialogue

    Zhengliang Shi, Weiwei Sun, Shuo Zhang, Zhen Zhang, Pengjie Ren, and Zhaochun Ren. Rade: Reference-assisted dialogue evaluation for open-domain dialogue. ArXiv, 2023. 15 Part I Appendix Contents A Practical Considerations and Societal Impact 16 A.1 Limitations . . . . . . . . ...

    2023

  88. [97]

    < THINK > , followed by a sub - query

  89. [98]

    < SEARCH > , followed by the citation ID

  90. [99]

    Please start with a special token ` < Final > ` followed by the final answer

    < RECORD > , followed by the answer to the sub - query Since this is a multi - hop question , your output should interleave the ` < THINK > ` , ` < SEARCH > ` , and ` < RECORD > ` tokens until you reach the final answer . Please start with a special token ` < Final > ` followe...

  91. [100]

    Godey ' s Lady ' s Book

    Arthur ' s Magazine | Arthur ' s Magazine (1844 -1846) was an American literary periodical published in Philadelphia in the 19 th century . Edited by T . S . Arthur , it featured works by Edgar A . Poe , J . H . Ingraham , Sarah Josepha Hale , Thomas G . Spear , and others . I...

  92. [101]

    Arthur ' s Magazine

    First for Women | First for Women is a women ' s magazine published by Bauer Media Group in the USA . The magazine was started in 1989. It is based in Englewood Cliffs , New Jersey . In 2011 , the circulation of the magazine was 1 ,310 ,696 copies . Your Output : < THINK > Whe...

    1989

  93. [102]

    Into Dust

    [3]... < RECORD > Hamilton , Ontario USER QUERY : Into Dust is a song by the alternative rock band formed in which city ? please give the final answer to the input question : Into Dust is a song by the alternative rock band formed in which city ? Hamilton , Ontario 39 Table 10...

Pith tools

Reviewed August 7, 2026 · model on record in the stance chip above.