←── back to feed
/topics/arxiv-cs-cl-papers-june-4-2026

arXiv cs.CL papers June 4 2026

50 items●1 sources●updated 13d ago●trend 0

Twenty papers posted to arXiv's cs.CL section on June 4, 2026 address challenges in language model reasoning, safety, and practical deployment. Topics span long-form generation, retrieval-augmented systems, AI-generated text detection, medical QA, memory management for conversational agents, and parameter-efficient fine-tuning.

  • POLARIS uses GRPO with frontier LLM judges and human-reference injection to improve small models' long-form creative writing quality and length consistency
  • Biomedical RAG study across 5 open-weight models (7B–72B), 10 QA datasets, and 4 retrieval methods finds only small, inconsistent improvements from retrieval
  • LazyAttention defers positional encoding in KV caching to improve reusability for long-context RAG and in-context learning without expensive re-encoding
  • LR-LoRA learns adapter rank during training instead of using fixed-rank constraint, enabling more flexible parameter-efficient fine-tuning
  • DOSEBENCH introduces 81 OTC dosing scenarios requiring dose-timing tracking and 24-hour intake computation to evaluate LLM safety in medical QA
[BLG]blog/rss50
POLARIS: Guiding Small Models to Write Long Stories
arXiv cs.CL · Rishanth Rajendhran, Jenna Russell, Mohit Iyyer, John Frederick Wieting · 13d
Discourse-Role Labels as Presentation-Time Variables for Context Use in Language Models
arXiv cs.CL · Jianguo Zhu · 13d
Computational conceptual history of scientific concepts: From early digital methods to LLMs
arXiv cs.CL · Michael Zichert, Arno Simons · 13d
SaliMory: Orchestrating Cognitive Memory for Conversational Agents
arXiv cs.CL · Kai Zhang, Xinyuan Zhang, Hongda Jiang, Shiun-Zu Kuo, Hyokun Yun, Ejaz Ahmed, Shereen Oraby, Ziyun Li, Sanat Sharma, Ann Lee, Ahmed A Aly, Anuj Kumar, Raffay Hamid, Xin Luna Dong · 13d
When Retrieval Doesn't Help: A Large-Scale Study of Biomedical RAG
arXiv cs.CL · Erfan Nourbakhsh, Rocky Slavin, Ke Yang, Anthony Rios · 13d
Expert-Aware Refusal Steering
arXiv cs.CL · Anna C. Marbut, Daniel R. Olson, Travis J. Wheeler · 13d
A Systematic Analysis of Linguistic Features in AI-Generated Text Detection Across Domains and Models
arXiv cs.CL · Yassir El Attar, Esra D\"onmez, Maximilian Maurer, Agnieszka Falenska · 13d
ACAT: A Collaborative Platform for Efficient Aspect-Based Sentiment Dataset Annotation
arXiv cs.CL · Ana-Maria Luisa Mocanu, Ciprian-Octavian Truica, Elena-Simona Apostol · 13d
Cross-Prompt Generalization in Detecting AI-Generated Fake News Using Interpretable Linguistic Features
arXiv cs.CL · Aya Vera-Jimenez, Samuel Jaeger, Calvin Ibenye, Dhrubajyoti Ghosh · 13d
MM-BizRAG: Rethinking Multimodal Retrieval-Augmented Generation for General Purpose Enterprise Q&A
arXiv cs.CL · Hanoz Bhathena, Parin Rajesh Jhaveri, Rohan Mittal, Prateek Singh, Aymen Kallala, Rachneet Kaur, Yiqiao Jin, Zhen Zeng, Adwait Ratnaparkhi, Denis Kochedykov · 13d
Supportive Token Revealing for Fast Diffusion Language Model Decoding
arXiv cs.CL · Giries Abu Ayoub, Mario Barbara, Llu\'is Pastor-P\'erez, Tanja Bien, Aneesh Barthakur, Alaa Maalouf, Loay Mualem · 13d
Can I Take Another Dose? Evaluating LLM Decision-Making Under Temporal Uncertainty in OTC Dosing QA
arXiv cs.CL · Maroof Kousar, Yibo Hu · 13d
Long Live Fine-Tuning: Task-Specific Transformers Outperform Zero-Shot LLMs for Misinformation Response Classification on Reddit
arXiv cs.CL · JooYoung Lee, Lin Tian, Angela Brillantes, Adriana-Simona Mih\u{a}i\c{t}\u{a}, Marian-Andrei Rizoiu · 13d
Using Text-Based Causal Inference to Disentangle Factors Influencing Online Review Ratings
arXiv cs.CL · Linsen Li, Aron Culotta, Nicholas Mattei · 13d
LazyAttention: Efficient Retrieval-Augmented Generation with Deferred Positional Encoding
arXiv cs.CL · Haocheng Xia, Mihir Pamnani, Hanxi Fang, Supawit Chockchowwat, Yongjoo Park · 13d
Parameter-Efficient Fine-Tuning with Learnable Rank
arXiv cs.CL · Arpit Garg, Simon Lucey, Hemanth Saratchandran · 13d
Noisy memory encoding explains negative polarity illusions
arXiv cs.CL · Yuhan Zhang, Edward Gibson · 13d
Deliberate Evolution: Agentic Reasoning for Sample-Efficient Symbolic Regression with LLMs
arXiv cs.CL · Xinyu Pang, Zhanke Zhou, Xuan Li, Fangrui Lv, Shanshan Wei, Sen Cui, Bo Han, Changshui Zhang · 13d
GlossAssist -- A Tool to Simplify Corpus Creation and Study the Effect of NLP Models in Low-Resource Documentation Settings
arXiv cs.CL · Bhargav Shandilya, Matt Buchholz, Alexis Palmer · 13d
DLLG: Dynamic Logit-Level Gating of LLM Experts
arXiv cs.CL · Bingnan Li, Zhaoyang Zhang, Xiaoze Liu, Yantao Shen, Shuli Jiang, Shuo Yang, Wei Xia, Zhuowen Tu, Stefano Soatto · 13d
When Clients Stop Following: A Cognitive Conceptualization Diagram-driven Framework for Strategic Counseling
arXiv cs.CL · Yihao Qin, Junyi Zhao, Changsheng Ma, Yongfeng Tao, Minqiang Yang, Chang Liu, Bin Hu · 13d
Read the Trace, Steer the Path: Trajectory-Aware Reinforcement Learning for Diffusion Language Models
arXiv cs.CL · Anant Khandelwal, Manish Gupta · 13d
MemoryDocDataSet: A Benchmark for Joint Conversational Memory and Long Document Reasoning
arXiv cs.CL · Qiyang Xie, Jialun Wu, Xinjie He, Su Liu, Shuai Xiao, Zhiyuan Lin, Weikai Zhou · 13d
Listening to the Workforce: Measuring Construction Worker Safety Attitudes from Social Media Discourse Using LLMs
arXiv cs.CL · Farouq Sammour, Yuxin Zhang, Zhenyu Zhang · 13d
Stepwise Reasoning Enhancement for LLMs via External Subgraph Generation
arXiv cs.CL · Xin Zhang, Yang Cao, Baoxing Wu, Kai Song, Siying Li · 13d
SePO: Self-Evolving Prompt Agent for System Prompt Optimization
arXiv cs.CL · Wangcheng Tao, Han Wu, Weng-Fai Wong · 13d
Learning What to Learn: Stage-Specific Data Sets for SFT-then-RL in Small Language Model Reasoning
arXiv cs.CL · Chongyang He, Rui Zhang, Zixuan Wang, Xin Li · 13d
Entity Binding Failures in Speech LLM Reasoning: Diagnosis and Chain-of-Thought Intervention
arXiv cs.CL · Ming-Hao Hsu, Xiaohai Tian, Jun Zhang, Zhizheng Wu · 13d
Off-Distribution Voices: Fanfiction Subgenres as Universal Vernacular Jailbreaks for Aligned LLMs
arXiv cs.CL · Zhongze Luo, Ruihe Shi, Zhenshuai Yin, Haoyue Liu, Weixuan Wan, Xiaoying Tang · 13d
SANE Schema-aware Natural-language Evaluation of Biological Data
arXiv cs.CL · Rolf Gattung, Martin Krueger, Markus Reischl · 13d
Self-Evolving Deep Research via Joint Generation and Evaluation
arXiv cs.CL · Han Zhu, Chengkun Cai, Yuanfeng Song, Xing Chen, Sirui Han, Yike Guo · 13d
SparDA: Sparse Decoupled Attention for Efficient Long-Context LLM Inference
arXiv cs.CL · Yaosheng Fu, Guangxuan Xiao, Xin Dong, Song Han, Oreste Villa · 13d
GENEB: Why Genomic Models Are Hard to Compare
arXiv cs.CL · Daria Ledneva, Mikhail Nuridinov, Denis Kuznetsov · 13d
Dynamic Infilling Anchors for Format-Constrained Generation in Diffusion Large Language Models
arXiv cs.CL · Boyan Han, Yiwei Wang, Yi Song, Yujun Cai, Chi Zhang · 13d
LDARNet: DNA Adaptive Representation Network with Learnable Tokenization for Genomic Modeling
arXiv cs.CL · Daria Ledneva, Denis Kuznetsov · 13d
Temporal Order Matters for Agentic Memory: Segment Trees for Long-Horizon Agents
arXiv cs.CL · Yifan Simon Liu, Liam Gallagher, Faeze Moradi Kalarde, Jiazhou Liang, Armin Toroghi, Scott Sanner · 13d
Cartridges at Scale: Training Modular KV Caches over Large Document Collections
arXiv cs.CL · Momchil Hardalov, Gonzalo Iglesias, Adri\`a de Gispert · 13d
VCIFBench: Evaluating Complex Instruction Following for Video Understanding
arXiv cs.CL · Huangchen Xu, Yuan Wu, Yi Chang · 13d
Fine-grained Fragment Retrieval in Multi-modal Long-form Dialogues
arXiv cs.CL · Hanbo Bi, Zhiqiang Yuan, Chongyang Li, Qiwei Yan, Zexi Jia, Jiapei Zhang, Xiaoyue Duan, Yingchao Feng, Jinchao Zhang, Jie Zhou · 13d
A Systematic Evaluation of Positional Bias in Multi-Video Summarization with MLLMs
arXiv cs.CL · Huangchen Xu, Yuan Wu, Yi Chang · 13d
Hybrid Adversarial Defence for Natural Language Understanding Tasks
arXiv cs.CL · Manar Abouzaid, Yang Wang, Chenghua Lin, Stuart E. Middleton · 13d
RAMPART: Registry-based Agentic Memory with Priority-Aware Runtime Transformation
arXiv cs.CL · Nikodem Tomczak · 13d
CYGNET: Cypher Gate for Neural Execution Triage and Cost Containment
arXiv cs.CL · Nikodem Tomczak · 13d
QO-Bench: Diagnosing Query-Operator-Preserving Retrieval over Typed Event Tuples
arXiv cs.CL · Mengao Zhang, Xiang Yang, Chang Liu, Tianhui Tan, Ke-wei Huang · 13d
LifeSide: Benchmarking Agents as Lifelong Digital Companions
arXiv cs.CL · Yuqian Wu, Zhijie Deng, Wei Chen, Junwei Li, Yutian Jiang, Junle Chen, Zhengjun Huang, Qingxiang Liu, Jing Tang, Jiaheng Wei, Yuxuan Liang · 13d
CRAFT: Cost-aware Refinement And Front-aware Tuning of Prompts
arXiv cs.CL · Shanu Kumar, Shubhanshu Khandelwal, Akhila Yesantarao Venkata, Parag Agrawal, Yova Kementchedjhieva, Manish Gupta · 13d
SMADE-IE: Sparse Multi-Agent Framework with Evidence-Driven Debate for Zero-Shot Information Extraction
arXiv cs.CL · Kenfeng Huang, Yi Cai, Xin Wu, Zikun Deng, Li Yuan · 13d
DuDi: Dual-Signal Distillation with Cross-Lingual Verbalizer
arXiv cs.CL · Patomporn Payoungkhamdee, Tinnakit Udsa, Jian Gang Ngui, Sarana Nutanong, Alham Fikri Aji, Peerat Limkonchotiwat · 13d
Rethinking Continual Experience Internalization for Self-Evolving LLM Agents
arXiv cs.CL · Jingwen Chen, Wenkai Yang, Shengda Fan, Wenbo Nie, Chenxing Sun, Shaodong Zheng, Yangen Hu, Lu Pan, Ke Zeng, Yankai Lin · 13d
Query-based Cross-Modal Projector Bolstering Mamba Multimodal LLM
arXiv cs.CL · SooHwan Eom, Jay Shim, Gwanhyeong Koo, Haebin Na, Mark A. Hasegawa-Johnson, Sungwoong Kim, Chang D. Yoo · 13d