Pith.

Artificial Intelligence · arXiv:2611.23702 v1 · submitted 18 Nov 2026 · 263 references

KnowMe Paper Radar: Lightweight Discovery of Promising Papers in the Age of Autoresearch

Zhuo Chen, Jian Wang, Jinjie Gu, Lijun Wu

The paper says

“The growing volume of topically similar papers makes it difficult to identify new work that can offer useful research insights or methodological guidance.”

more of the abstract

We present KnowMe Paper Radar, a lightweight discovery tool that combines multiple sources of evidence into personalized reading priorities. Users specify topics, priorities, exclusions and dates; the system brings together topical relevance, recency, author research histories, collaboration paths and direction-specific institutional context. By exposing each score component and its supporting sources, KnowMe helps readers select promising papers for their current questions and discover relevant work beyond familiar authors. Automatic acquisition, version merging and graph-cache reuse keep this workflow convenient without large language model (LLM) inference during ordinary use. Experiments on 174,640 acquired records show more than threefold faster repeated context preparation, while cached abstracts improve normalized discounted cumulative gain at rank 10 (NDCG@10) from 0.412 to 0.598 on expert-judged retrieval tasks. Complementary analyses examine how the same accumulated records support semantic retrieval, research-history modeling and literature-grounded planning. A browser interface and a structured command-line interface (CLI) make these capabilities available to both researchers and research agents, positioning KnowMe as a practical upstream tool for literature-informed recursive self-improvement (RSI) workflows.

from the abstract

Pith says

Machine review · deepseek-v4-flash · 19 Nov 2026 · read the full text (285,153 characters) · not checked by a person

It can be wrong. Challenge this review or reply to it.

Holds up, with conditions. A well-engineered discovery tool that combines author histories, collaboration paths, and institution context into a lightweight, LLM-free paper ranking system, with solid efficiency measurements and a careful retrieval evaluation—though the central claim about identifying 'promising' papers is not validated against future impact. This line is wrong

10 minor

How often objections from reviews like this one are right: not yet measured

Are you an author? What you can do.

The paper, first character to last

The reviewer was given the whole text, 285,153 characters. Not checked: figures, code, what the 263 references say.

10 minor objections

The paper says, at Abstract

Pith says: minor objection 1 of 10

The NDCG@10 values are reported as 0.412 and 0.598, but in Table 4 (Table 4 in the paper) the values are reported as 41.22 and 59.80 (multiplied by 100). The abstract should use consistent scaling or explicitly state the multiplier.

The paper says, at §2.3, Equation 1

Pith says: minor objection 2 of 10

The term Δ_m(p,u) for manual preferences is introduced but never defined or discussed in detail. It would be helpful to briefly explain what kinds of manual preferences are included and how they are set.

show 8 more

The paper says, at §2.4, Equation 2

Pith says: minor objection 3 of 10

The notation 'd matches k' is ambiguous. It would be clearer to specify what kind of matching is used (e.g., keyword, semantic, or categorical).

The paper says, at §2.5, Equation 4

Pith says: minor objection 4 of 10

The restart probability 0.65 and the resulting weights 0.2275 and 0.1225 are stated without justification. It would be helpful to explain how these values were chosen or whether they were tuned.

The paper says, at §5.3, Table 4 (Table 4 in the paper is the facet retrieval table)

Pith says: minor objection 5 of 10

The SPECTER2 results are reported with a specific base checkpoint. It would be helpful to specify which SPECTER2 variant was used (e.g., SPECTER2-base) and whether any task-specific adapters were applied.

The paper says, at §5.4, Figure 6

Pith says: minor objection 6 of 10

The paper reports the paired change as 18.59 points, but the absolute values are 0.412 and 0.598, which is a difference of 0.186. The paper should clarify that the values are multiplied by 100 for readability.

The paper says, at §6.1, Table 10 (Table 10 in the paper is the planner table)

Pith says: minor objection 7 of 10

The specificity and feasibility differences are small and have large standard deviations. The paper should discuss whether these differences are practically meaningful, especially given the small sample size (16 directions).

The paper says, at Appendix A, Figure 1 (Figure 1 in the paper is about arXiv growth)

Pith says: minor objection 8 of 10

The growth projections are based on a model selected by holdout WAPE, but the projections extend to 2035, far beyond the validation horizon. The paper acknowledges this limitation, but it would be helpful to also report prediction intervals for the projections.

The paper says, at §2.8, Table 1 (Table 1 in the paper is about reuse levels)

Pith says: minor objection 9 of 10

The paper reports that 1,533 IDs were recovered from cached responses. It would be helpful to know what fraction of these recovered IDs were actually relevant to the expanded scope (i.e., how many were false positives).

The paper says, at §5.5, Figure 8 (Figure 8 in the paper is about topic dynamics)

Pith says: minor objection 10 of 10

The indexing makes it difficult to compare the absolute levels of paper share and author participation. It would be helpful to also report the raw counts or proportions.

What would settle it

Recompute Table 2 row 3 with the corrected normalization in §A.2; if the entry shifts by more than 10% the headline claim weakens.

from the skeptic's reading

How the review went

5 passes by deepseek-v4-flash · 19 Nov 2026

  1. Reader holds up, with conditions confidence moderate
  2. Skeptic agrees 8 s · 78,546 tokens read, 354 written
  3. Referee 10 minor 35 s · 78,903 tokens read, 3,576 written
  4. Check against the paper kept 0 of 4 objections 42 s · 159,270 tokens read, 5,000 written
  5. Editor wrote the verdict above 11 s · 79,203 tokens read, 560 written

Not run on this paper: the simulated authors' defence.

Pith's passes raised 4 objections. A check against the paper's text kept 0 and withdrew 4 (4 Pith's reviewer misread the paper).

The review in full

The review in brief, machine-written

This paper presents KnowMe Paper Radar, a practical tool for filtering the growing arXiv stream. Its main contribution is a decomposed scoring model that combines topical relevance, recency, author research histories, collaboration paths, and institution context into a single, inspectable ranking, all without LLM inference during ordinary use. The system is well-engineered: source-graph reuse reduces preparation time by more than threefold, and cached abstracts improve NDCG@10 from 0.412 to 0.598 on the CSFCube expert-judged retrieval benchmark. The efficiency measurements are paired and reproducible, and the retrieval evaluation uses an external benchmark with careful protocol.

The paper is transparent about its components building on prior work (AMiner, OpenAlex, SPECTER2). The evaluation is thorough in several dimensions: acquisition cost, preparation time, retrieval quality, author coverage, and institution context. The reading-list experiments show that the added author and institution context changes selection at fixed budgets (e.g., 14.7% replacement at 1,000 reads), and the literature-to-plan handoff study demonstrates that full-context packets expose more traceable research histories in downstream agent plans.

The soft spot is the central framing. The paper claims to identify 'promising' papers, but the evaluation does not measure future impact. The reading-list experiments show that author and institution context does not change topical relevance much (model assessments show close average readouts between full context and baseline), and the paper does not ship code or a commit hash for full reproducibility. The comparison with other personalized discovery tools is qualitative. These gaps are acknowledged in the paper's limitations section, but they mean the core value proposition—that personal context helps select papers that will be useful—is not directly validated.

Who this is for: researchers who want a practical, inspectable tool for daily arXiv triage, and developers building research-agent workflows that need a lightweight literature-discovery component. The paper deserves a serious referee: the engineering is solid, the evaluation is careful, and the tool addresses a real problem. The main recommendation would be to either (a) reframe the claims to match what is actually measured (topical relevance + efficiency), or (b) add a prospective study of whether the selected papers become influential. As is, the paper is a useful systems contribution with a slightly overclaimed framing.

The strongest objection

The paper presents a well-engineered discovery tool with solid efficiency measurements and a careful retrieval evaluation on an external benchmark. The NDCG improvement and threefold speedup are credible. However, the central framing of 'promising papers' is not validated against future impact: the reading-list experiments show that the added author and institution context does not change topical relevance much, and the paper does not ship code or a commit hash. The comparison with other personalized discovery tools is qualitative. These gaps warrant a conditional acceptance rather than full acceptance. No formal verification is provided, and correctness risk is medium because the flagship claim is under-supported by the evaluation design.

The reader's weakest_assumption correctly identifies that the core value of the personal-context ranking is not established because the evaluation uses a retrieval benchmark and model-based reading assessments, not a prospective study of whether selected papers become influential. The paper's own reading-list experiments (Figure 9c,d) show that the added author and institution context does not change topical relevance much, and the paper does not ship code or a commit hash. The comparison with other personalized discovery tools is qualitative. These gaps warrant a conditional acceptance rather than full acceptance. No formal verification is provided, and correctness risk is medium because the flagship claim is under-supported by the evaluation design.

The referee's report

Not published. Pith's referee wrote this assessment before its objections were checked against the paper, and 4 of its 4 major objections did not survive that check.

Pith's reading of the paper (a machine's summary, not the authors')

The paper presents KnowMe Paper Radar, a lightweight discovery tool that helps researchers identify promising new papers by combining topical relevance with author research histories, collaboration paths, and institutional context. The system is designed to work without LLM inference during ordinary use, relying instead on cached abstracts and a reusable source graph. Experiments on 174,640 acquired records show that cached abstracts improve retrieval NDCG@10 from 0.412 to 0.598 on expert-judged tasks, and source-graph reuse reduces preparation time by more than threefold. The tool provides both a browser interface and a CLI for research agents, positioning it as an upstream literature source for recursive self-improvement workflows.

What carries the argument. The central mechanism is a decomposed scoring model (Equation 1) that combines a base content score with personal author fit, institutional context, and manual preferences, each with bounded influence. The author fit uses a leave-author-out reference with shrinkage (Equations 2-5), and the institutional context combines a sourced prior with graph evidence from a two-step random walk with restart (Equations 6-8). The system separates source acquisition and graph preparation from subsequent reading requests, enabling reuse without LLM inference.

Core claim. The paper claims that a practical, LLM-free discovery tool can effectively combine topical relevance with author histories, collaboration paths, and institutional context to help researchers select promising new papers. The key design choices are: (1) a decomposed scoring model that separates content, author, institution, and recency components, each with bounded influence; (2) automatic acquisition and version merging that caches abstracts and preserves a research graph across changing interests; and (3) source-graph reuse that reduces repeated preparation time by more than threefold. The evaluation shows that cached abstracts improve NDCG@10 from 0.412 to 0.598 on expert-judged retrieval tasks, and that the accumulated records support semantic retrieval, graph learning, and literature-grounded planning without requiring LLM inference during daily triage.

The premise it rests on. The paper assumes that 'promising' papers can be identified from topical relevance plus author histories and institutional context, without measuring future impact. The evaluation uses a retrieval benchmark and model-based reading assessments, not a prospective study of whether selected papers become influential.

If the paper is right (Pith's inferences, not the authors')

If the approach is correct, researchers can maintain a personal literature library that becomes more useful over time without requiring expensive LLM inference for each query.

The reusable source graph could serve as a shared infrastructure for multiple research agents, reducing redundant literature acquisition across different research questions.

The decomposed scoring model makes reading priorities interpretable, allowing researchers to understand why a paper was recommended and to adjust their preferences accordingly.

The system's ability to surface authors beyond the initial catalogue through graph propagation could help researchers discover relevant work from unfamiliar groups.

The CLI interface enables integration with automated research workflows, potentially supporting recursive self-improvement systems that need to select literature without human intervention.

The paper's emphasis on LLM-free operation suggests a design philosophy where literature discovery should remain computationally accessible even as LLM-based research tools become more powerful. This could be important for maintaining equitable access to research infrastructure.

The evaluation focuses on retrieval quality and system cost rather than on whether the selected papers actually lead to better research outcomes. A prospective study tracking whether papers selected by KnowMe become influential would strengthen the claim that the tool identifies 'promising' work.

The institutional context component, which combines a static prior with dynamic graph evidence, could be extended to incorporate real-time updates from funding announcements, conference programs, or preprint server trends, making it more responsive to emerging research directions.

The paper's framing around 'autoresearch' and RSI workflows suggests that the authors envision a future where literature discovery is increasingly automated, but the tool's design deliberately keeps the human researcher in the loop through inspectable scores and source attribution.

Check this review

What was read

The passes were given all 285,153 characters of the paper's LaTeX source.

Fingerprint of that text (SHA-256): ff9844c4a88e9e390c303f0b625abeaa0a691ea8dbbdc848f736cc6ea812f34a

Pith does not keep that text. Rebuilt from the paper's source or PDF by the same code, it has this fingerprint if it is the same text. From a copy of Pith's code:

PYTHONPATH=. python scripts/review_input_check.py 2611.23702

The passes, in the order they ran

  1. Reader deepseek-v4-flash · medium effort · $0.021 · instructions of the Reader, fingerprint f9ead7300382…
  2. Skeptic deepseek-v4-flash · medium effort · $0.0072 · instructions of the Skeptic, fingerprint 8f88c0f61488…
  3. Referee deepseek-v4-flash · medium effort · $0.0078 · instructions of the Referee, fingerprint a852d99aea0d…
  4. Check against the paper deepseek-v4-flash · medium effort · $0.0096 · instructions of the Check against the paper, fingerprint 6c7a6883d551…
  5. Editor deepseek-v4-flash · medium effort · $0.0073 · instructions of the Editor, fingerprint e7211f6a3d27…
  6. Circularity check deepseek-v4-flash · medium effort · $0.0072 · instructions of the Circularity check, fingerprint e9946e39d67b…
  7. Summary for the page deepseek-v4-flash · medium effort · temperature 0.2 · $0.0076 · instructions of the Summary for the page, fingerprint e4f56bf568e7…

The passes cost $0.068 in model calls together.

Version stamped on this review: pith-pipeline@v1.4.1-daily-deepseek-or-medium+deepseek/deepseek-v4-flash+p.d60f54226f00

Each link opens a pass's instructions as plain text, the exact bytes its fingerprint was taken over. Every stored set of instructions.

curl -s https://pith.science/methodology/prompts/f9ead730038216984cfb56d7cb34332bdf691828296a1de4245f3d6667d4bf56 | sha256sum

prints the fingerprint of the reader's instructions. The same works for every pass.

The receipt

The review's fingerprint was taken on 19 Nov 2026. The review you are reading still matches it.

Fingerprint of the review (SHA-256): 64ee58b4de587019c97c1352282640f6157e3451661763ecd2e234ed0cce0720

curl -s 'https://pith.science/api/v1/verdicts/2611.23702?form=canonical' | sha256sum

prints the fingerprint of the review as it stands now. The rule, in four steps.

A matching fingerprint shows the review has not changed since that date. It does not show that the review is right.

How often reviews like this one are right

Reviews like this one: written by deepseek-v4-flash, from the paper's LaTeX source.

How often objections from reviews like this one are right: not yet measured

Every value above, in the review's public record (JSON)

Reviews by people

This review was written by a machine. No person has reviewed this paper on Pith yet. A scientist listed on Pith can review it for a fee, and their review is published here whatever it concludes.

Request a human review

How Pith checks papers · The full audit of this paper

The paper

Abstract

The growing volume of topically similar papers makes it difficult to identify new work that can offer useful research insights or methodological guidance. We present KnowMe Paper Radar, a lightweight discovery tool that combines multiple sources of evidence into personalized reading priorities. Users specify topics, priorities, exclusions and dates; the system brings together topical relevance, recency, author research histories, collaboration paths and direction-specific institutional context. By exposing each score component and its supporting sources, KnowMe helps readers select promising papers for their current questions and discover relevant work beyond familiar authors. Automatic acquisition, version merging and graph-cache reuse keep this workflow convenient without large language model (LLM) inference during ordinary use. Experiments on 174,640 acquired records show more than threefold faster repeated context preparation, while cached abstracts improve normalized discounted cumulative gain at rank 10 (NDCG@10) from 0.412 to 0.598 on expert-judged retrieval tasks. Complementary analyses examine how the same accumulated records support semantic retrieval, research-history modeling and literature-grounded planning. A browser interface and a structured command-line interface (CLI) make these capabilities available to both researchers and research agents, positioning KnowMe as a practical upstream tool for literature-informed recursive self-improvement (RSI) workflows.

The paper's own sentences, arranged, kept beside the review and separate from it. Each label is drawn from a closed vocabulary; the number beside it is the probability the classifier gave that label. No sentence in this record was written by a model.

abstract compiled · 2026-11-19

The authors' main claim, as the classifier picked it out 0.56

“Experiments on 174,640 acquired records show more than threefold faster repeated context preparation, while cached abstracts improve normalized discounted cumulative gain at rank 10 (NDCG@10) from 0.412 to 0.598 on expert-judged retrieval tasks.”

How the main claim is stated: asserted 1.00

What the abstract itself states, as probabilities: claims first 0.04 · states limitation 0.16 · human participants 0.30 · controlled comparison 0.33 · randomized assignment 0.01 · states quantitative result 0.99 · compares with prior results 0.51 · mentions code or data availability 0.11

The abstract, sentence by sentence

This table labels the abstract only (arXiv listing abstract v1). The number is the probability the classifier gave the label.

SentenceRoleThe authors' own workStates the main conclusionStrength
The growing volume of topically similar papers makes it difficult to identify new work that can offer useful research insights or methodological guidance. background 0.96 0.05 0.03 asserted 1.00
We present KnowMe Paper Radar, a lightweight discovery tool that combines multiple sources of evidence into personalized reading priorities. contribution 1.00 0.88 0.19 asserted 1.00
Users specify topics, priorities, exclusions and dates; the system brings together topical relevance, recency, author research histories, collaboration paths and direction-specific institutional context. method 0.93 0.59 0.07 asserted 1.00
By exposing each score component and its supporting sources, KnowMe helps readers select promising papers for their current questions and discover relevant work beyond familiar authors. contribution 0.58 0.57 0.27 asserted 0.99
Automatic acquisition, version merging and graph-cache reuse keep this workflow convenient without large language model (LLM) inference during ordinary use. method 0.94 0.78 0.06 asserted 1.00
Experiments on 174,640 acquired records show more than threefold faster repeated context preparation, while cached abstracts improve normalized discounted cumulative gain at rank 10 (NDCG@10) from 0.412 to 0.598 on expert-judged retrieval tasks. result 1.00 0.96 0.28 asserted 1.00
Complementary analyses examine how the same accumulated records support semantic retrieval, research-history modeling and literature-grounded planning. contribution 0.72 0.86 0.07 asserted 0.96
A browser interface and a structured command-line interface (CLI) make these capabilities available to both researchers and research agents, positioning KnowMe as a practical upstream tool for literature-informed recursive self-improvement (RSI) workflows. contribution 0.60 0.74 0.20 asserted 0.93

What kind of paper the classifier reports this is: method or tool 1.00 · evidence from observation or experiment 0.81

What was examined, and what was not

Sources read

  • arXiv listing abstract v1

Not examined

  • full text
  • figures and tables
  • references and citing sentences
  • relations to other papers

Levels run

  • P abstract

Not run

  • S sections
  • T body sentences
  • C citations
  • R cross-paper relations

How the labels were made

Compiled by classifier-1.13.0 via TypeSafe API, question bank abstract-p1-20260923, 2026-11-19: 1 request, 6,386 input tokens.

Every quoted sentence on this record is the authors'. Every label is from Pith's closed vocabulary, with the probability shown beside it. This record as JSON

References (263)
  1. [1] 2026 , howpublished = on Pith
  2. [2] Yue Huang and Zhangchen Xu and Yuchen Ma and Wenjie Wang and Zheyuan Liu and Ziwei Xu and others , title =. CoRR , volume = on Pith
  3. [3] Yu, Lezhi and Xu, Xiaogang and Zhou, Yuhua and He, Shuibing and Pan, Aimin , title =. CoRR , volume = on Pith
  4. [4] Oh, Yerim and Lee, Young-Jun and Ahn, Jaewoo and Kim, Gunhee and Kang, Dongyeop , title =. CoRR , volume = on Pith
  5. [5] Xuehai Wang and Haowei Qin and Tongxin Liu and Junkai Li and Buqiang Xu and Jintian Zhang and others , title =. CoRR , volume = on Pith
  6. [6] Huaiyu Wan and Yutao Zhang and Jing Zhang and Jie Tang , title =. Data Intell. , volume = on Pith
  7. [7] Pierre Azoulay and Alessandro Bonatti and Joshua L. Krieger , title =. Res. Policy , volume = on Pith
  8. [8] Susan Feng Lu and Ginger Zhe Jin and Brian Uzzi and Benjamin Jones , title =. Sci. Rep. , volume = on Pith
  9. [9] Uthman Jinadu and Parsa Ghazvinian and Anjila Budathoki and Benjamin M. Ampel and Rajshekhar Sunderraman and Yi Ding , title =. EMNLP , pages =. 2026 , publisher = on Pith
  10. [10] Jason Priem and Heather A. Piwowar and Richard Orr , title =. CoRR , volume = on Pith
  11. [11] Markus Flicke and Glenn Angrabeit and Madhav Iyengar and Vitalii Protsenko and Illia Shakun and Jovan Cicvaric and others , title = on Pith
  12. [12] Fuqiang Wang and Song Tan and Zheng Guo and Jiaohao Fu and Xinglong Xu and Bihui Yu and others , title =. CoRR , volume = on Pith
  13. [13] Junshu Pan and Luodan Zhang and Yifeng Lu and Mengfan Zhao and Ming Luo and Zijie Yang and Yue Zhang and Rui Shang , title =. CoRR , volume = on Pith
  14. [14] Shuofei Qiao and Yunxiang Wei and Jiazheng Fan and Bin Wu and Busheng Zhang and Mengru Wang and others , title =. CoRR , volume = on Pith
  15. [15] Shuofei Qiao and Yunxiang Wei and Xuehai Wang and Bin Wu and Boyang Xue and Ningyu Zhang and others , title =. ICML , pages =. 2026 , volume = on Pith
  16. [16] Mengru Wang and Junfeng Fang and Shuofei Qiao and Zhenqian Xu and Haoming Xu and Haoxiong Wang and others , title =. CoRR , volume = on Pith
  17. [17] Zhisong Qiu and Kangqi Song and Shengwei Tang and Shuofei Qiao and Lei Liang and Huajun Chen and Shumin Deng , title =. CoRR , volume = on Pith
  18. [18] Hu, Weihua and Fey, Matthias and Zitnik, Marinka and Dong, Yuxiao and Ren, Hongyu and Liu, Bowen and Catasta, Michele and Leskovec, Jure , title =. NeurIPS , pages =. 2020 , volume = on Pith
  19. [19] Amanpreet Singh and Mike D'Arcy and Arman Cohan and Doug Downey and Sergey Feldman , title =. EMNLP , pages =. 2023 , publisher = on Pith
  20. [20] Weiwei Sun and Lingyong Yan and Xinyu Ma and Shuaiqiang Wang and Pengjie Ren and Zhumin Chen and Dawei Yin and Zhaochun Ren , title =. EMNLP , pages =. 2023 , publisher = on Pith
show 80 more
  1. [21] Chen Liu , title =. 2026 , howpublished = on Pith
  2. [22] Dingjie Song and Hanrong Zhang and Dawei Liu and Yixin Liu and Zongxia Li and Zhengqing Yuan and others , title =. EMNLP (Demos) , pages =. 2026 , publisher = on Pith
  3. [23] Atsuyuki Miyai and Mashiro Toyooka and Takashi Otonari and Zaiying Zhao and Kiyoharu Aizawa , title =. CoRR , volume = on Pith
  4. [24] Ruofeng Yang and Yongcan Li and Shuai Li , title =. CoRR , volume = on Pith
  5. [25] Weitong Qian and Beicheng Xu and Zhongao Xie and Bowen Fan and Guozheng Tang and Jiale Chen and others , title =. CoRR , volume = on Pith
  6. [26] Jiefeng Chen and Bhavana Dalvi Mishra and Jaehyun Nam and Rui Meng and Tomas Pfister and Jinsung Yoon , title =. ICML , pages =. 2026 , publisher = on Pith
  7. [27] Chenglei Si and Zitong Yang and Yejin Choi and Emmanuel J. Cand. CoRR , volume = on Pith
  8. [28] Mingming Zhao and Jiqian Dong and Kangping Xu and Zadid Hasan and Chengrui Fan and Shan Jiang and others , title =. CoRR , volume = on Pith
  9. [29] Amy Xin and Jiening Siow and Junjie Wang and Zijun Yao and Fanjin Zhang and Jian Song and Lei Hou and Juanzi Li , title =. CoRR , volume = on Pith
  10. [30] Yuvraj Virk and Zack Edds and Chunqiu Steven Xia and Lingming Zhang , title =. CoRR , volume = on Pith
  11. [31] Yougang Lyu and Xi Zhang and Xinhao Yi and Yuyue Zhao and Shuyu Guo and Wenxiang Hu and others , title =. CoRR , volume = on Pith
  12. [32] Lange, Robert Tjarko and Grillotti, Luca and Yamada, Yutaro and Charakorn, Rujikorn and Ha, David , title =. CoRR , volume = on Pith
  13. [33] Jiaqi Liu and Shi Qiu and Mairui Li and Bingzhou Li and Haonian Ji and Siwei Han and others , title =. CoRR , volume = on Pith
  14. [34] Bobo Li and Hao Fei and Tianjie Ju and Mong-Li Lee and Wynne Hsu , title =. CoRR , volume = on Pith
  15. [35] Li, Ming and Li, Dai and Ning, Xuying and Sun, Bo and Li, Rui and Zhang, Yi and others , title =. CoRR , volume = on Pith
  16. [36] Xiyuan Yang and Sheikh Sarwar and Jingru Cheng and Zhan Shi and Duanshun Li and Huiyuan Chen and others , title =. CoRR , volume = on Pith
  17. [37] Wang, Zijun and Liu, Zewen and Lin, Minhua and Weng, Zhaotian and Shi, Zhan and He, Bing and others , title =. CoRR , volume = on Pith
  18. [38] Haozhe Liu and Tian Ye and Sensen Gao and Qihang Cao and Yitong Li and Mingchen Zhuge and others , title =. CoRR , volume = on Pith
  19. [39] Guiyao Tie and Jiawen Shi and Dingjie Song and Yixiao Huang and Ziji Sheng and Xueyang Zhou and others , title =. CoRR , volume = on Pith
  20. [40] Aniketh Garikaparthi and Manasi Patwardhan and Arman Cohan , title =. CoRR , volume = on Pith
  21. [41] Alisia Maria Lupidi and Bhavul Gauri and Thomas Foster and Bassel Al Omari and Despoina Magka and Alberto Pepe and others , title =. CoRR , volume = on Pith
  22. [42] Wanghan Xu and Shuo Li and Tianlin Ye and Qinglong Cao and Yixin Chen and Hengjian Gao and others , title =. CoRR , volume = on Pith
  23. [43] Jiayu Wang and Weijiang Lv and Bowen Fu and Jing Fu and Jiayi Song and Lingyu Zhang and others , title =. CoRR , volume = on Pith
  24. [44] Jintian Zhang and Kewei Xu and Jingsheng Zheng and Zhuoyun Yu and Yuqi Zhu and Yujie Luo and others , title =. ICLR , year = on Pith
  25. [45] Shitanshu Bhushan and Yunxiang Zhang and Lu Wang , title =. CoRR , volume = on Pith
  26. [46] Zhang, Yudi and Cao, Mingyu and Yin, Lu and Pechenizkiy, Mykola and Liu, Shiwei , title =. CoRR , volume = on Pith
  27. [47] Fanqing Meng and Lingxiao Du and Qiguang Chen and Ziqi Zhao and Haocheng Lu and Mengkang Hu and Michael Qizhe Shieh , title =. CoRR , volume = on Pith
  28. [48] Liu, Yuxin and Wang, Yuxuan and Lei, Zhenxin and Meng, Lingchen and Sun, Yuchong and Lin, Junming and others , title =. CoRR , volume = on Pith
  29. [49] Chi, Yizhe and Li, Wenyi and Hong, Deyao and Wang, Xiaoqiu and Gao, Mingju and Yang, Kaisen and others , title =. CoRR , volume = on Pith
  30. [50] Bansal, Aayam and Balaji, Keertan , title =. CoRR , volume = on Pith
  31. [51] Fei, Yanlin and Liu, Nazhou and Yu, Xinmiao and Chen, Shaolong and Li, Lei and Thapa, Rahul and others , title =. CoRR , volume = on Pith
  32. [52] Sun, Youran and Ren, Xingyu and Yi, Chugang and Zhang, Kejia and Wu, Di and Cao, Yingkang and others , title =. CoRR , volume = on Pith
  33. [53] Ning, Jingjie and Li, Xueqi and Kong, Yibo and Li, Dongting , title =. CoRR , volume = on Pith
  34. [54] Zhang, Kate and Li, Yuante , title =. CoRR , volume = on Pith
  35. [55] Zhibo Yang and Chen Zhang and Yuewei Zhang and Hao Wang , title =. CoRR , volume = on Pith
  36. [56] Chuhan Shi and Xiaoquan Ren and Sicheng Song and Haobo Li and Rui Sheng and Yushi Sun , title =. CoRR , volume = on Pith
  37. [57] Guoqiang Zhang and Kexin Tan and Ming Zhang and Li Ju and Wenqing Jing and Zhonghan Yue and others , title =. CoRR , volume = on Pith
  38. [58] JiaJun Hou and Hexuan Deng and Wenxiang Jiao and Xuebo Liu and Xiaopeng Ke and Derek F. Wong and Min Zhang , title =. EMNLP (Findings) , pages =. 2026 , publisher = on Pith
  39. [59] Sy-Tuyen Ho and Minghui Liu and Huy Nghiem and Furong Huang , title =. CoRR , volume = on Pith
  40. [60] Yiling Ma and Yilun Zhao and Sihong Wu and Manasi Patwardhan and Arman Cohan , title =. CoRR , volume = on Pith
  41. [61] Bai, Jiamu and Zhang, Lizhu and Yu, Xin and Wu, Yanhong and Wang, Zellux and Li, Serena and others , title =. CoRR , volume = on Pith
  42. [62] Jingsheng Zheng and Jintian Zhang and Yujie Luo and Yuren Mao and Yunjun Gao and Lun Du and Huajun Chen and Ningyu Zhang , title =. ACL , pages =. 2026 , publisher = on Pith
  43. [63] Swati Rajwal and Sanjay Das and Tirthankar Ghosal , title =. CoRR , volume = on Pith
  44. [64] Varun Gumma and Navonil Majumder and Soumitra Sinhahajari and Soujanya Poria , title =. CoRR , volume = on Pith
  45. [65] Yuchen Wang and Zhongzhi Luan , title =. CoRR , volume = on Pith
  46. [66] Fengxian Ji and Yuke Li and Jingpu Yang and Juanfan Wu and Fan Zhang and Zhexuan Cui and others , title =. CoRR , volume = on Pith
  47. [67] Xiaopeng Li and Wenlin Zhang and Yingyi Zhang and Pengyue Jia and Yejing Wang and Yichao Wang and others , title =. SIGIR , pages = on Pith
  48. [68] Guo, Fang and Zhu, Qi and Pei, Rongcan and He, Shuqi and Chen, Hui and Zhang, Yue , title =. CoRR , volume = on Pith
  49. [69] Nishant Balepur and Malachi Hamada and Varsha Kishore and Sergey Feldman and Amanpreet Singh and Pao Siangliulue and others , title =. ACL , pages =. 2026 , publisher = on Pith
  50. [70] Rima Hazra and Sayan Layek and Somnath Banerjee and Soumen Chakrabarti and Animesh Mukherjee , title =. CoRR , volume = on Pith
  51. [71] Maayan Sharon and Tom Hope , title =. CoRR , volume = on Pith
  52. [72] Chenyang Shao and Yong Li and Fengli Xu , title =. CoRR , volume = on Pith
  53. [73] Xingyu Liu and Yu Dong and Qizhen Yu and Shiyu Cheng and Zhe Wang and Guan Li and others , title =. CoRR , volume = on Pith
  54. [74] Yuyang Huang and Bobo Li and JiaJia Song and Yuzhe Ding and Chong Teng and Fei Li and Donghong Ji , title =. EMNLP (Findings) , pages =. 2026 , publisher = on Pith
  55. [75] Yuwen Du and Tian Jin and Jing Kang and Xianghe Pang and Jingyi Chai and Tingjia Miao and others , title =. CoRR , volume = on Pith
  56. [76] Gaurav Sahu and Laurent Charlin and Christopher Pal , title =. CoRR , volume = on Pith
  57. [77] Peng Xia and Rujun Han and Zifeng Wang and Yanfei Chen and Yufan Zhuang and Yoonho Lee and others , title =. CoRR , volume = on Pith
  58. [78] Siwei Wu and Jincheng Ren and Yizhi Li and Haau-Sing Li and Chengran Yang and Yuxuan Zhang and others , title =. CoRR , volume = on Pith
  59. [79] Sungho Park and Wonjoong Kim and Rongyuan Tan and Jue Zhang and Wook-Shin Han and Pengfei Gao and others , title =. CoRR , volume = on Pith
  60. [80] Park, Sungho and Kim, Wonjoong and Zhang, Jue and Han, Wook-Shin and Gao, Pengfei and Park, Chanyoung and others , title =. CoRR , volume = on Pith
  61. [81] Haoyu Dong and Yuhang Zhou and Zihao Lin and Yifan Wu and Bo Peng and Mingyi Wang and others , title =. CoRR , volume = on Pith
  62. [82] Xinke Jiang and Tao Feng and Weixuan Xu and Zhixin Zhang and Zhibang Yang and Wentao Zhang and others , title =. CoRR , volume = on Pith
  63. [83] Cehao Yang and Xiaojun Wu and Xueyuan Lin and Chengjin Xu and Xuhui Jiang and Hui Xiong and Jian Guo , title =. CoRR , volume = on Pith
  64. [84] Dhruv Srikanth and Bingchen Zhao and Dixing Xu and Yuxiang Wu and Zhengyao Jiang , title =. CoRR , volume = on Pith
  65. [85] Yujie Luo and Xiangyuan Ru and Jingsheng Zheng and Jingjing Wang and Yuqi Zhu and Jintian Zhang and others , title =. EMNLP , pages =. 2026 , publisher = on Pith
  66. [86] Yixuan Tang and Yi Yang , title =. CoRR , volume = on Pith
  67. [87] Leyao Wang and Yanan He and Peng Chen and Asaf Yehudai and Yixin Liu and Rex Ying and Michal Shmueli-Scheuer and Arman Cohan , title =. CoRR , volume = on Pith
  68. [88] Au Kwok Chun and Abhigyan Acherjee and Amrutha Rao and Zaiqian Chen and Kazem Meidani and C. Bayan Bruss and Micah Goldblum , title =. CoRR , volume = on Pith
  69. [89] Jiazhen Ji and Shouhong Ding , title =. CoRR , volume = on Pith
  70. [90] Nguyen, Michael and Tan, Wei Chen and Hassan, Nurul Aisyah and Raman, Arvind and Lim, Li Hua and Razak, Ahmad Faiz , title =. CoRR , volume = on Pith
  71. [91] Chengyin Hu and Dingyi Lu and Jiaju Han and Xiang Chen and Weiwen Shi and Jiahuan Long and Yiwei Wei and Jiujiang Guo , title =. CoRR , volume = on Pith
  72. [92] Atindra Jha and Margaret Li and Jure Leskovec and Percy Liang and Luke Zettlemoyer , title =. CoRR , volume = on Pith
  73. [93] Xu Zou and Jie Tang , title =. CoRR , volume = on Pith
  74. [94] Yuqing Zhou and Hong Wang and Manqing Mao and Zhuoer Wang and Samson Koelle and Jie Yuan and others , title =. CoRR , volume = on Pith
  75. [95] Jie Sun and Mao Zheng and Mingyang Song and Zeyuan Liu and Gengsheng Li and Houcheng Jiang and others , title =. CoRR , volume = on Pith
  76. [96] Gengsheng Li and Mao Zheng and Mingyang Song and Jie Sun and Zeyuan Liu and Ruiqi Liu and others , title =. CoRR , volume = on Pith
  77. [97] Jiaxin Bai and Yue Guo and Yifei Dong and Jiaxuan Xiong and Tianshi Zheng and Yixia Li and others , title =. CoRR , volume = on Pith
  78. [98] Tianshi Zheng and Rui Wang and Xiyun Li and Kelvin Kiu Wai Tam and Newt Nguyen Kim Hue Nam and Wei Fan and Yangqiu Song and Tianqing Fang , title =. CoRR , volume = on Pith
  79. [99] Guibin Zhang and Leo Lu and Fangzhou Xie and Kang Zhu and Junhao Wang and Zhifei Xie and others , title =. CoRR , volume = on Pith
  80. [100] Yanze Wang and Pengfei Yao and Tianyi Sun and Chuanrui Hu and Yan Xiao and Xiaotian Luo and others , title =. CoRR , volume = on Pith

The first 100 of 263 references are listed here. All of them are in this paper's citation record.

Challenges and discussion

Disagree with the review, or know something it missed? Leave a comment, or challenge this review formally: a challenge becomes a permanent entry in this paper's record, and its decision is published here.

No comments yet.

Sign in to comment