Proposes linearized 2-simplicial attention using random features for O(n) complexity, combined with Kimi Delta Attention, achieving state-of-the-art accuracy on long sequences.

Aritra Das, Dhruman Gupta, Debayan Gupta

This paper analyzes the validity of privileged likelihood as token credit, proposing three checks to validate its usefulness in on-policy self-distillation.

Xuan-Phi Nguyen, Shrey Pandit, Yiran Zhao et al.

UniMoMo compresses trained sparse MoE recommendation models via expert merging, achieving speedups of up to 2.21× with minimal performance loss.

Lei Xin, Bin Gu, Peize Li et al.

FailForge extracts procedural skills from failed instances to improve code agent success rate by 6.6 points on SWE benchmarks.

Dongyi Lv, Fushun E, Aichen Cai et al.

Introduces IoA-Suite for hidden disagreement detection; model F1 only 49.5%, private context is bottleneck.

Kaiming Liu, Fuwen Luo, Ziyue Wang et al.

Unified on-policy self-distillation (USD) dynamically adjusts token weights and privileged information via a single dual variable, boosting reasoning performance.

Yongkang Yang, Zhezheng Hao, Hong Zhang et al.

SkillCDG employs a two-layer graph model for long SKILL compliance detection, achieving 12.8% F1 improvement and reducing token usage by 64.3%.

Shuaitao Zhao, Feng Ni, Lichao Ma et al.

Proposes an attribution framework based on representation, selection, and reasoning, empirically analyzing failure modes in multi-page visually-rich document understanding.

Lewei Xu, Yihao Ding, Zihan Xu et al.

Vision Language Models reduce input tokens by encoding time-series as 2D plots, improving accuracy and reducing energy consumption.

Bhavika Jalli, Nikhil Korati Prasanna, Jayanta Choudhury

Proposes CoBa, a compute-balanced routing method reducing 49.1% tokens, achieving 85.13% accuracy in test-time reasoning.

Yan Zhou, Yue Ouyang, Kaiyang Zheng et al.

FinRank leverages manually curated hard negatives to enhance financial QA evidence discrimination, with a benchmark of 1185 question-answer pairs.

Sasan Mansouri, Daniel Saad, Mark Wahrenburg et al.

WNM-3D adds 3D scene tokens to closed-loop VLN; the supplied paper text reports gains on GN-Bench but omits exact scores.

Yuehao Huang, Yunzi Wu, Xiaotao Zhang et al.

MemWM enhances text-based world models with memory, boosting SSF by up to 206.3% and success rate by 65.4% in multi-environment benchmarks.

Yujun Wang, Tao Zhang, Jinhe Bi et al.

SkillEval learns interpretable linear directions in model embeddings to evaluate four skill document dimensions, correlating with downstream task performance.

Jiahui Han, Qinuo Li, Ziheng Peng et al.

bioMoR enhances genomic learning efficiency via biology-guided recursion mixture, boosting macro-F1 by 8.2%.

Koushik Howlader, Tirtho Roy, Md Tauhidul Islam et al.

Proposes SYF, an LLM-based interactive recommendation system enabling multimodal real-time content control.

Ziyun Xu, Bosen Ding, Yue Zhang et al.

This paper analyzes challenges in evaluating explanation methods for static and evolving data, proposing bias detection and concept drift explanations with experimental validation.

Jerzy Stefanowski

HarnessOpt-Bench benchmark evaluates 5 frontier LLMs in costly, stochastic harness optimization, revealing significant model differences and room for improvement.

Varun Ursekar, Apaar Shanker, Yash Maurya et al.

CAV and TopK SAE auditing show that concept recoverability is not equivalent to influence on speaking scores.

Arya Labroo, Mengjie Qian, Kate Knill

AgentOPSD employs recursive Bayesian belief updates for turn-level credit assignment, outperforming baselines with 89.1% success on ALFWorld.

Zi-Han Wang, Zhengxi Lu, Zhiyuan Yao et al.