Introduces LFQA-HP-1M dataset with nine scoring metrics, achieving model performance comparable to human preferences in long-form QA evaluation.

Rafid Ishrak Jahan, Fahmid Shahriar Iqbal, Sagnik Ray Choudhury

Enhanced persuasive agents using cross-disciplinary strategies, achieving higher success rates, especially for low-intent users.

Shinnosuke Nozue, Yuto Nakano, Yotaro Watanabe et al.

LARFT combines length-aware reinforcement learning and hindsight length awareness, improving length instruction following by +20.92 points on average.

Wei Zhang, Lintong Du, Yuanhe Zhang et al.

This study evaluates 8 state-of-the-art LLMs' calibration in predicting causal discovery algorithm performance, revealing widespread 'algorithmic blindness'.

Sohan Venkatesh, Ashish Mahendran Kurapath, Tejas Melkote

Proposed TSD-KD improves reasoning by combining preference-based indirect and confidence-based direct distillation, outperforming baselines by up to 54.4%.

Minsang Kim, Seung Jun Baek

SibylSense adapts rubric generation via memory tuning and adversarial probing, enhancing RL performance.

Yifei Xu, Guilherme Potje, Shivam Shandilya et al.

Yor-Sarc is the first gold-standard dataset for sarcasm detection in Yorùbá, with 436 annotated instances and a Fleiss' κ of 0.766.

Toheeb Aduramomi Jimoh, Tabea De Wille, Nikola S. Nikolov

VIRAASAT enhances Indian cultural reasoning with Symbolic Chain-of-Manipulation, achieving up to 20% improvement.

Harshul Raj Surana, Arijit Maji, Aryan Vats et al.

Introduced X-Value benchmark to assess cross-lingual values judgment in LLMs with 4,750 QA pairs.

Yukun Chen, Xinyu Zhang, Boyi Deng et al.

MemoryArena evaluates agent memory in multi-session tasks, revealing gaps in current assessments.

Zexue He, Yu Wang, Churan Zhi et al.

Proposes Query-as-Anchor framework with hierarchical encoders and joint contrastive-autoregressive training for scenario-adaptive user representations, achieving state-of-the-art results.

Jiahao Yuan, Yike Xu, Jinyong Wen et al.

Proposes knowledge profiling revealing near-saturation encoding but persistent recall bottleneck; thinking improves recall significantly.

Nitay Calderon, Eyal Ben-David, Zorik Gekhman et al.

Proposes PACED-RL, leveraging the partition function as an online accuracy signal to improve sample efficiency in distribution-matching training of LLMs.

Dohyung Kim, Minbeom Kim, Jeonghye Kim et al.

Proposes On-Policy Context Distillation (OPCD), minimizing reverse KL divergence on self-generated trajectories, boosting knowledge internalization for tasks like math reasoning and transfer.

Tianzhu Ye, Li Dong, Xun Wu et al.

Few-Step Diffusion Language Models via Trajectory Self-Distillation improve reasoning tasks.

Tunyu Zhang, Xinxi Zhang, Ligong Han et al.

Proposes Composition-RL, automatically combining prompts to improve reasoning, achieving up to 21.4% pass@1 gains on 30B models.

Xin Xu, Clive Bai, Kai Yang et al.

Introduces GRU-Mem for long-context reasoning, achieving up to 400% efficiency improvement by addressing memory explosion and lack of exit mechanisms.

Leheng Sheng, Yongtao Zhang, Wenchang Ma et al.

LEMUR creates a multilingual EU law corpus and fine-tunes state-of-the-art embedding models with contrastive learning to improve legal retrieval accuracy.

Narges Baba Ahmadi, Jan Strich, Martin Semmann et al.

LycheeMemory enables long-context reasoning via compressed memory, reducing GPU usage by 2x and speeding inference by 6x.

Zhuoen Chen, Dongfang Li, Meishan Zhang et al.

Document reconstruction enables scalable long-context RLVR, improving RULER and LongBench v2 performance.

Yao Xiao, Lei Wang, Yue Deng et al.