pith. sign in

arxiv: 2606.04155 · v1 · pith:K4CWRRERnew · submitted 2026-06-02 · 💻 cs.HC · cs.CL· cs.CY

SocialCoach: Personalized Social Skill Learning with RL-based Agentic Tutoring and Practice

Tianfu Wang , Max Xiong , Jianxun Lian , Hongyuan Zhu , Zhengyu Hu , Yuxuan Lei , Linxiao Gong , Xiaofang Li
show 3 more authors
Peiting Tsai Nicholas Jing Yuan Qi Zhang
This is my paper

Pith reviewed 2026-06-28 08:02 UTC · model grok-4.3

classification 💻 cs.HC cs.CLcs.CY
keywords social skill learningLLM agentic tutoringreinforcement learning schedulingpersonalized educationsoft skills trainingimmersive practiceknowledge corpus constructionadaptive learning paths
0
0 comments X

The pith

SocialCoach builds an LLM agentic system that constructs expert social-skill corpora and uses RL in learner simulations to schedule personalized practice paths.

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

The paper presents SocialCoach as a system that automatically assembles a theory-to-practice knowledge base from expert sources via multi-agent pipelines. It then personalizes training through a prescription-retrieval-adaptation loop whose policy is trained by reinforcement learning inside a simulated learner environment to solve cold-start problems. Immersive goal-driven practice, causality-based assessment, and reflective tutoring are layered on top to close the knowing-doing gap for skills such as negotiation and leadership. Experiments report gains in simulated pathway quality and external-judge tutoring ratings relative to baselines, together with early deployment feedback indicating high user engagement.

Core claim

SocialCoach demonstrates that an integrated pipeline of multi-agent corpus construction, RL-optimized adaptive scheduling in a learner simulation, and LLM-driven immersive practice with reflective feedback produces higher-quality learning pathways and tutoring interactions than non-RL or non-agentic baselines while delivering usable personalization at scale.

What carries the argument

The adaptive practice scheduling module that executes a prescription-retrieval-adaptation process whose policy is optimized by reinforcement learning inside a simulated learner environment to maximize long-term experience and handle cold-start.

If this is right

  • Personalized social-skill curricula can be generated and scheduled without continuous human expert availability.
  • Reinforcement learning policies trained in simulation can overcome cold-start limitations for new learners.
  • Combining immersive practice with causality-driven assessment narrows the gap between knowing social techniques and applying them.
  • The same architecture supplies a template for other soft-skill domains that currently lack scalable coaching.

Where Pith is reading between the lines

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

  • If the simulation-to-reality transfer holds, the same RL scheduling loop could be reused across different tutoring domains with only the knowledge corpus swapped.
  • Real-time user telemetry could be fed back to refine the simulation model, creating a closed-loop improvement cycle not described in the current deployment.
  • Extending the multi-agent corpus pipeline to ingest live expert videos or case studies could further reduce reliance on static sources.

Load-bearing premise

The simulated learner environment used to train the reinforcement-learning scheduler accurately models real human learning dynamics and transfers to actual users.

What would settle it

A randomized controlled trial that measures real-user skill gains (via pre/post performance in negotiation or leadership tasks) after SocialCoach tutoring versus a matched non-RL baseline and reports no statistically significant difference.

Figures

Figures reproduced from arXiv: 2606.04155 by Hongyuan Zhu, Jianxun Lian, Linxiao Gong, Max Xiong, Nicholas Jing Yuan, Peiting Tsai, Qi Zhang, Tianfu Wang, Xiaofang Li, Yuxuan Lei, Zhengyu Hu.

Figure 1
Figure 1. Figure 1: The social skills training landscape. (a) Examples [PITH_FULL_IMAGE:figures/full_fig_p001_1.png] view at source ↗
Figure 2
Figure 2. Figure 2: An overview of the SocialCoach Framework for personalized social skill training at scale. [PITH_FULL_IMAGE:figures/full_fig_p003_2.png] view at source ↗
Figure 3
Figure 3. Figure 3: The proposed theory-to-practice social knowledge [PITH_FULL_IMAGE:figures/full_fig_p004_3.png] view at source ↗
Figure 4
Figure 4. Figure 4: Scenario reference retriever optimization with re [PITH_FULL_IMAGE:figures/full_fig_p005_4.png] view at source ↗
Figure 5
Figure 5. Figure 5: Evaluation results on tutoring guidance. [PITH_FULL_IMAGE:figures/full_fig_p008_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: Distribution of scenario contextual types. [PITH_FULL_IMAGE:figures/full_fig_p011_6.png] view at source ↗
Figure 9
Figure 9. Figure 9: Representative learner profile distribution. [PITH_FULL_IMAGE:figures/full_fig_p012_9.png] view at source ↗
Figure 8
Figure 8. Figure 8: Top 10 social skills by frequency. Workplace (27%), with coverage of other areas like public and edu￾cational settings. This balance suggests the corpus provides con￾textually relevant scenarios for a range of real-world interpersonal challenges. B.2 Expert Likert Ratings of Corpus Tag Annotation Consistency. To assess the reliability of the auto￾mated tagging process, we measure the label agreement of hum… view at source ↗
Figure 10
Figure 10. Figure 10: An illustrative learning journey case. B.5 Case Study of Learning Journey We illustrate a case study in [PITH_FULL_IMAGE:figures/full_fig_p013_10.png] view at source ↗
Figure 11
Figure 11. Figure 11: Several interfaces of our product, EQoach. [PITH_FULL_IMAGE:figures/full_fig_p014_11.png] view at source ↗
Figure 12
Figure 12. Figure 12: Questionnaire results from 50 participants. [PITH_FULL_IMAGE:figures/full_fig_p014_12.png] view at source ↗
read the original abstract

Social skills such as negotiation and leadership are crucial for personal and professional success in today's interconnected world. However, scalable and effective training remains a significant challenge due to the scarcity of expert coaching. In this paper, we introduce SocialCoach, a holistic LLM-powered agentic tutoring system for personalized social skill development at scale. First, SocialCoach automatically constructs a pedagogically-grounded, theory-to-practice knowledge corpus from diverse expert sources, leveraging a multi-agent pipeline. Second, to personalize the learning journey, it employs an adaptive practice scheduling module that follows a prescription-retrieval-adaptation process. To maximize the long-term learning experience while overcoming the cold-start problem, this policy is optimized within a learner simulation environment through reinforcement learning. Finally, SocialCoach integrates immersive, goal-driven practice, causality-driven proficiency assessment and knowledge-grounded, reflective tutoring to help address the knowing-doing gap. We deploy it in our product, EQoach, and conduct extensive experiments. The results show that SocialCoach improves simulated pathway quality and judge-rated tutoring quality over baseline approaches, while early user feedback indicates strong perceived engagement and usefulness. These findings suggest a practical architecture for personalized and gamified pedagogical platforms on soft skill learning.

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

2 major / 2 minor

Summary. The paper introduces SocialCoach, an LLM-powered agentic system for scalable personalized social skill training. It constructs a pedagogically-grounded knowledge corpus via a multi-agent pipeline from expert sources, employs an RL-optimized adaptive scheduling policy (prescription-retrieval-adaptation loop with cold-start handling) inside a learner simulation environment, and integrates immersive goal-driven practice, causality-driven assessment, and reflective tutoring. Deployed in EQoach, it claims improvements in simulated pathway quality and judge-rated tutoring quality over baselines, plus positive early user feedback on engagement and usefulness.

Significance. If the simulation-to-real transfer holds, the architecture offers a concrete, deployable approach to addressing expert-coach scarcity in soft-skill domains through automated corpus construction and long-horizon RL scheduling. The integration of theory-to-practice corpus building with RL personalization and immersive practice is a practical contribution to HCI and educational technology. The lack of demonstrated correlation between simulation metrics and real-user outcomes, however, limits the strength of the claims.

major comments (2)
  1. [§3.2 (adaptive scheduling) and Experiments] The central claim that the RL policy improves simulated pathway quality and enables practical personalization rests on the learner simulation environment accurately modeling human learning dynamics and transfer. No section provides evidence (e.g., correlation studies or real-user validation experiments) that simulation metrics predict proficiency gains, engagement, or transfer outside the simulation.
  2. [Experiments / Results] The abstract and results statements claim performance improvements in simulated pathway quality and judge-rated tutoring quality, yet supply no quantitative metrics, explicit baseline definitions, effect sizes, or statistical tests. This prevents assessment of whether the reported gains are robust or merely artifacts of the simulation setup.
minor comments (2)
  1. [§3.1] The multi-agent pipeline for corpus construction is described at a high level; adding pseudocode or a diagram of the agent roles and information flow would improve reproducibility.
  2. [§3.2] Implementation details for the RL reward function, state representation, and cold-start handling mechanism are referenced but not fully specified, making it difficult to replicate the scheduling policy.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive and detailed feedback. We address each major comment below with clarifications on the scope of our evaluation and specific plans for revision.

read point-by-point responses

  1. Referee: [§3.2 (adaptive scheduling) and Experiments] The central claim that the RL policy improves simulated pathway quality and enables practical personalization rests on the learner simulation environment accurately modeling human learning dynamics and transfer. No section provides evidence (e.g., correlation studies or real-user validation experiments) that simulation metrics predict proficiency gains, engagement, or transfer outside the simulation.

    Authors: We acknowledge that the current manuscript does not include correlation studies or direct real-user validation experiments linking simulation metrics to real-world proficiency gains or transfer. The learner simulation is constructed from established pedagogical principles in social skill learning literature and expert input, with additional grounding provided by judge-rated tutoring quality and early deployment feedback in EQoach. We agree this represents a limitation for claims about practical personalization. In revision we will add an explicit Limitations section discussing the simulation-to-real gap and outlining planned future validation studies. revision: partial

  2. Referee: [Experiments / Results] The abstract and results statements claim performance improvements in simulated pathway quality and judge-rated tutoring quality, yet supply no quantitative metrics, explicit baseline definitions, effect sizes, or statistical tests. This prevents assessment of whether the reported gains are robust or merely artifacts of the simulation setup.

    Authors: We agree that the abstract and high-level results statements would benefit from greater quantitative transparency. Although comparative results appear in the experiments section, we will revise the abstract and results summary to include explicit quantitative metrics (pathway quality scores and tutoring quality ratings), clearly define all baselines (e.g., random scheduling and non-RL policies), and report effect sizes together with statistical tests. This change will make the robustness of the gains directly assessable. revision: yes

Circularity Check

1 steps flagged

RL-optimized scheduling policy shows gains in the same simulation used for training

specific steps
  1. fitted input called prediction [Abstract]
    "To maximize the long-term learning experience while overcoming the cold-start problem, this policy is optimized within a learner simulation environment through reinforcement learning. [...] The results show that SocialCoach improves simulated pathway quality and judge-rated tutoring quality over baseline approaches"

    The policy is trained inside the simulation to maximize long-term learning experience; the claimed improvement in 'simulated pathway quality' is therefore the training outcome by construction rather than a held-out prediction or external validation of the simulation's fidelity.

full rationale

The paper optimizes its adaptive scheduling policy via RL inside a learner simulation to maximize long-term learning experience, then reports that SocialCoach 'improves simulated pathway quality' over baselines. This reported improvement is the direct result of the optimization objective rather than an independent test. No external real-user proficiency correlation is shown to break the loop. The judge-rated and user-feedback results are downstream and do not validate the simulation itself. This matches the fitted_input_called_prediction pattern but does not extend to the full architecture or other claims.

Axiom & Free-Parameter Ledger

0 free parameters · 0 axioms · 0 invented entities

Abstract-only review supplies no concrete free parameters, axioms, or invented entities; all entries left empty.

pith-pipeline@v0.9.1-grok · 5782 in / 1087 out tokens · 14229 ms · 2026-06-28T08:02:40.760985+00:00 · methodology

discussion (0)

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

Reference graph

Works this paper leans on

51 extracted references · 18 canonical work pages · 6 internal anchors

  1. [1]

    Taylor N Allbright, Julie A Marsh, Kate E Kennedy, Heather J Hough, and Susan McKibben. 2019. Social-emotional learning practices: Insights from outlier schools.Journal of Research in Innovative Teaching & Learning12, 1 (2019), 35–52

    2019

  2. [2]

    Mitchell

    Jonathan Bassen, Bharathan Balaji, Michael Schaarschmidt, Candace Thille, Jay Painter, Dawn Zimmaro, Alex Games, Ethan Fast, and John C. Mitchell. 2020. Reinforcement Learning for the Adaptive Scheduling of Educational Activities. InCHI

    2020

  3. [3]

    Jaime Carbonell and Jade Goldstein. 1998. The use of MMR, diversity-based reranking for reordering documents and producing summaries. InSIGIR

    1998

  4. [4]

    Jiangjie Chen, Xintao Wang, Rui Xu, Siyu Yuan, Yikai Zhang, Wei Shi, Jian Xie, Shuang Li, Ruihan Yang, Tinghui Zhu, et al. 2024. From persona to personaliza- tion: A survey on role-playing language agents.arXiv preprint arXiv:2404.18231 (2024)

    work page arXiv 2024

  5. [5]

    Zhendong Chu, Shen Wang, Jian Xie, Tinghui Zhu, Yibo Yan, Jinheng Ye, Aoxiao Zhong, Xuming Hu, Jing Liang, Philip S Yu, et al. 2025. Llm agents for education: Advances and applications.arXiv preprint arXiv:2503.11733(2025)

    work page arXiv 2025

  6. [6]

    Jiwon Chun, Yankun Zhao, Hanlin Chen, and Meng Xia. 2025. PlanGlow: Person- alized Study Planning with an Explainable and Controllable LLM-Driven System. InL@S

    2025

  7. [7]

    2017.Soft skills needed for the 21st century workforce

    Susan A Dean. 2017.Soft skills needed for the 21st century workforce. Walden University

    2017

  8. [8]

    DeepSeek-AI. 2025. DeepSeek-R1-0528 Release. https://api-docs.deepseek.com/ news/news250528

    2025

  9. [9]

    David Dinucu-Jianu, Jakub Macina, Nico Daheim, Ido Hakimi, Iryna Gurevych, and Mrinmaya Sachan. 2025. From Problem-Solving to Teaching Problem- Solving: Aligning LLMs with Pedagogy using Reinforcement Learning.arXiv preprint arXiv:2505.15607(2025)

    work page arXiv 2025

  10. [10]

    2015.Handbook of social and emotional learning: Research and practice

    Joseph A Durlak. 2015.Handbook of social and emotional learning: Research and practice. Guilford Publications

    2015

  11. [11]

    Julian G Elliott, Steven E Stemler, Robert J Sternberg, Elena L Grigorenko, and Newman Hoffman. 2011. The socially skilled teacher and the development of tacit knowledge.British Educational Research Journal37, 1 (2011), 83–103

    2011

  12. [12]

    Anna Fang, Hriday Chhabria, Alekhya Maram, and Haiyi Zhu. 2025. Social Simulation for Everyday Self-Care: Design Insights from Leveraging VR, AR, and LLMs for Practicing Stress Relief. InCHI

    2025

  13. [13]

    Tao Ge, Xin Chan, Xiaoyang Wang, Dian Yu, Haitao Mi, and Dong Yu. 2024. Scaling synthetic data creation with 1,000,000,000 personas.arXiv preprint arXiv:2406.20094(2024)

    work page internal anchor Pith review Pith/arXiv arXiv 2024

  14. [14]

    Frank M. Gresham. 2015.Disruptive Behavior Disorders: Evidence-Based Practice for Assessment and Intervention. Guilford Publications

    2015

  15. [15]

    Hengnian Gu, Zhiyi Duan, Pan Xie, and Dongdai Zhou. 2024. Modeling balanced explicit and implicit relations with contrastive learning for knowledge concept recommendation in MOOCs. InWWW

    2024

  16. [16]

    Jiawei Gu, Xuhui Jiang, Zhichao Shi, Hexiang Tan, Xuehao Zhai, Chengjin Xu, Wei Li, Yinghan Shen, Shengjie Ma, Honghao Liu, et al . 2024. A survey on llm-as-a-judge.arXiv preprint arXiv:2411.15594(2024)

    work page internal anchor Pith review Pith/arXiv arXiv 2024

  17. [17]

    Michael Guevarra, Indronil Bhattacharjee, Srijita Das, Christabel Wayllace, Car- rie Demmans Epp, Matthew E Taylor, and Alan Tay. 2025. An LLM-Guided Tutoring System for Social Skills Training. InAAAI

    2025

  18. [18]

    Sarah Susanna Hoppler, Robin Segerer, and Jana Nikitin. 2022. The six com- ponents of social interactions: Actor, partner, relation, activities, context, and evaluation.Frontiers in Psychology12 (2022), 743074

    2022

  19. [19]

    Tiancheng Hu, Joachim Baumann, Lorenzo Lupo, Nigel Collier, Dirk Hovy, and Paul Röttger. 2025. Simbench: Benchmarking the ability of large language models to simulate human behaviors.arXiv preprint arXiv:2510.17516(2025)

    work page internal anchor Pith review Pith/arXiv arXiv 2025

  20. [20]

    Chris Davis Jaldi, Eleni Ilkou, Noah Schroeder, and Cogan Shimizu. 2025. Ed- ucation in the era of Neurosymbolic AI.Journal of Web Semantics85 (2025), 100857

    2025

  21. [21]

    Stephanie M Jones and Emily J Doolittle. 2017. Social and emotional learning: Introducing the issue.The future of children27, 1 (2017), 3–11

    2017

  22. [22]

    Matthew Jörke, Shardul Sapkota, Lyndsea Warkenthien, Niklas Vainio, Paul Schmiedmayer, Emma Brunskill, and James A. Landay. 2025. GPTCoach: Towards LLM-Based Physical Activity Coaching. InCHI

    2025

  23. [23]

    Harold W. Kuhn. 1955. The Hungarian method for the assignment problem. Naval Research Logistics Quarterly2, 1-2 (1955), 83–97

    1955

  24. [24]

    Qingyao Li, Wei Xia, Li’ang Yin, Jiarui Jin, and Yong Yu. 2024. Privileged knowl- edge state distillation for reinforcement learning-based educational path recom- mendation. InKDD

    2024

  25. [25]

    Jiayu Liu, Zhenya Huang, Tong Xiao, Jing Sha, Jinze Wu, Qi Liu, Shijin Wang, and Enhong Chen. 2024. SocraticLM: Exploring socratic personalized teaching with large language models.Advances in Neural Information Processing Systems 37 (2024), 85693–85721

    2024

  26. [26]

    Yujie Luo, Xiangyuan Ru, Kangwei Liu, Lin Yuan, Mengshu Sun, Ningyu Zhang, Lei Liang, Zhiqiang Zhang, Jun Zhou, Lanning Wei, Da Zheng, Haofen Wang, and Huajun Chen. 2025. OneKE: A Dockerized Schema-Guided LLM Agent-based Knowledge Extraction System. InWWW

    2025

  27. [27]

    Setareh Maghsudi, Andrew Lan, Jie Xu, and Mihaela van Der Schaar. 2021. Personalized education in the artificial intelligence era: what to expect next. IEEE Signal Processing Magazine38, 3 (2021), 37–50

    2021

  28. [28]

    OpenAI. 2025. GPT-5 System Card. https://cdn.openai.com/gpt-5-system-card. pdf. Published August 7, 2025

    2025

  29. [29]

    getting us through:

    Sarah M Ovink and Brian D Veazey. 2011. More than “getting us through:” A case study in cultural capital enrichment of underrepresented minority under- graduates.Research in higher education52, 4 (2011), 370–394

    2011

  30. [30]

    Jiahuan Pei, Fanghua Ye, Xin Sun, Wentao Deng, Koen Hindriks, and Junxiao Wang. 2025. Conversational Education at Scale: A Multi-LLM Agent Workflow for Procedural Learning and Pedagogic Quality Assessment. InEMNLP

    2025

  31. [31]

    Erasmo Purificato, Ludovico Boratto, and Ernesto William De Luca. 2024. User modeling and user profiling: A comprehensive survey.arXiv preprint arXiv:2402.09660(2024)

    work page arXiv 2024

  32. [32]

    Cheng Qian, Zuxin Liu, Akshara Prabhakar, Jielin Qiu, Zhiwei Liu, Haolin Chen, Shirley Kokane, Heng Ji, Weiran Yao, Shelby Heinecke, et al . 2025. UserRL: Training Interactive User-Centric Agent via Reinforcement Learning.arXiv preprint arXiv:2509.19736(2025)

    work page arXiv 2025

  33. [33]

    Sajratul Y Rubaiat and Hasan M Jamil. 2025. Context-Aware Scientific Knowledge Extraction on Linked Open Data using Large Language Models.arXiv preprint arXiv:2506.17580(2025)

    work page arXiv 2025

  34. [34]

    Surbhi Seema Sethi and Kanishk Jain. 2024. AI technologies for social emotional learning: recent research and future directions.Journal of Research in Innovative Teaching & Learning17, 2 (2024), 213–225

    2024

  35. [35]

    Bernstein

    Omar Shaikh, Valentino Emil Chai, Michele Gelfand, Diyi Yang, and Michael S. Bernstein. 2024. Rehearsal: Simulating Conflict to Teach Conflict Resolution. In CHI

    2024

  36. [36]

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

    work page internal anchor Pith review Pith/arXiv arXiv 2024

  37. [37]

    Lee S. Shulman. 1998. Theory, practice, and the education of professionals.The Elementary School Journal98, 5 (1998), 511–526

    1998

  38. [38]

    Ian Steenstra, Farnaz Nouraei, and Timothy W Bickmore. 2025. Scaffolding Empathy: Training Counselors with Simulated Patients and Utterance-level Performance Visualizations. InCHI

    2025

  39. [39]

    Tianlong Wang, Xianfeng Jiao, Yinghao Zhu, Zhongzhi Chen, Yifan He, Xu Chu, Junyi Gao, Yasha Wang, and Liantao Ma. 2025. Adaptive activation steering: A tuning-free llm truthfulness improvement method for diverse hallucinations categories. InWWW. Tianfu Wang et al

    2025

  40. [40]

    Tianfu Wang, Yi Zhan, Jianxun Lian, Zhengyu Hu, Nicholas Jing Yuan, Qi Zhang, Xing Xie, and Hui Xiong. 2025. LLM-powered Multi-agent Framework for Goal- oriented Learning in Intelligent Tutoring System. InWWW

    2025

  41. [41]

    Weixun Wang, Shaopan Xiong, Gengru Chen, Wei Gao, Sheng Guo, Yancheng He, Ju Huang, Jiaheng Liu, Zhendong Li, Xiaoyang Li, et al. 2025. Reinforcement Learning Optimization for Large-Scale Learning: An Efficient and User-Friendly Scaling Library.arXiv preprint arXiv:2506.06122(2025)

    work page arXiv 2025

  42. [42]

    Shirley Wu, Michel Galley, Baolin Peng, Hao Cheng, Gavin Li, Yao Dou, Weixin Cai, James Zou, Jure Leskovec, and Jianfeng Gao. 2025. CollabLLM: From Passive Responders to Active Collaborators. InICML

    2025

  43. [43]

    An Yang, Anfeng Li, Baosong Yang, Beichen Zhang, Binyuan Hui, Bo Zheng, Bowen Yu, Chang Gao, Chengen Huang, Chenxu Lv, et al. 2025. Qwen3 technical report.arXiv preprint arXiv:2505.09388(2025)

    work page internal anchor Pith review Pith/arXiv arXiv 2025

  44. [44]

    Diyi Yang, Caleb Ziems, William Held, Omar Shaikh, Michael S Bernstein, and John Mitchell. 2024. Social skill training with large language models.arXiv preprint arXiv:2404.04204(2024)

    work page arXiv 2024

  45. [45]

    Haofei Yu, Zhengyang Qi, Yining Zhao, Kolby Nottingham, Keyang Xuan, Bod- hisattwa Prasad Majumder, Hao Zhu, Paul Pu Liang, and Jiaxuan You. 2025. Sotopia-RL: Reward Design for Social Intelligence.arXiv preprint arXiv:2508.03905 (2025)

    work page arXiv 2025

  46. [46]

    Yi Zhan, Qi Liu, Weibo Gao, Zheng Zhang, Tianfu Wang, Shuanghong Shen, Junyu Lu, and Zhenya Huang. 2025. CoderAgent: Simulating Student Behavior for Personalized Programming Learning with Large Language Models. InIJCAI

    2025

  47. [47]

    Xinnong Zhang, Jiayu Lin, Xinyi Mou, Shiyue Yang, Xiawei Liu, Libo Sun, Hanjia Lyu, Yihang Yang, Weihong Qi, Yue Chen, et al. 2025. Socioverse: A world model for social simulation powered by llm agents and a pool of 10 million real-world users.arXiv preprint arXiv:2504.10157(2025)

    work page arXiv 2025

  48. [48]

    Yanzhao Zhang, Mingxin Li, Dingkun Long, Xin Zhang, Huan Lin, Baosong Yang, Pengjun Xie, An Yang, Dayiheng Liu, Junyang Lin, et al . 2025. Qwen3 Embedding: Advancing Text Embedding and Reranking Through Foundation Models.arXiv preprint arXiv:2506.05176(2025)

    work page internal anchor Pith review Pith/arXiv arXiv 2025

  49. [49]

    Weixiang Zhao, Xingyu Sui, Yulin Hu, Jiahe Guo, Haixiao Liu, Biye Li, Yanyan Zhao, Bing Qin, and Ting Liu. 2025. Teaching Language Models to Evolve with Users: Dynamic Profile Modeling for Personalized Alignment.arXiv preprint arXiv:2505.15456(2025)

    work page arXiv 2025

  50. [50]

    Jialun Zhong, Wei Shen, Yanzeng Li, Songyang Gao, Hua Lu, Yicheng Chen, Yang Zhang, Wei Zhou, Jinjie Gu, and Lei Zou. 2025. A comprehensive survey of reward models: Taxonomy, applications, challenges, and future.arXiv preprint arXiv:2504.12328(2025)

    work page arXiv 2025

  51. [51]

    prescription

    Xuhui Zhou, Hao Zhu, Leena Mathur, Ruohong Zhang, Haofei Yu, Zhengyang Qi, Louis-Philippe Morency, Yonatan Bisk, Daniel Fried, Graham Neubig, et al. 2024. SOTOPIA: Interactive Evaluation for Social Intelligence in Language Agents. In ICLR. A Details of Framework Design A.1 Knowledge Categorization Details To ensure our knowledge corpus is pedagogically so...

    2024