The paper introduces DAN, a training‑free inference‑time framework that improves affective reasoning in multimodal large language models. It combines a Hierarchical Emotional Reasoning Chain (HERC) to better capture fine‑grained visual cues and a Contrastive Discriminative Visual Pruning (CDVP) module to isolate discriminative tokens for semantically similar emotions. Experiments show significant gains, notably a +10.47% improvement on the WebEmo25 benchmark with Qwen3‑VL‑8B‑Instruct.
By Cheng Ye, Weidong Chen, Zhaobo Qi, Beier Zhu, Zhendong Mao
The paper introduces RareTrap, a framework that estimates the probability of severe behaviors in black‑box large language models. RareTrap constructs a geometry‑aware mapping from a low‑dimensional latent space into token‑embedding space using a surrogate LLM, creating an explicit and reproducible distribution over input prompts. By applying a response‑level performance function and sequential rare‑event simulation, RareTrap concentrates evaluations on increasingly severe behaviors while preserving probability, enabling estimation of such behaviors with as few as 200 evaluations across multiple open‑weight and frontier models.
By Elsayed Eshra, Ali Al-Lawati, Dongwon Lee, Suhang Wang
The paper shows that pretrained transformers often stop using their depth early, following only a few lines of context. A small rank‑8 LoRA applied to an early layer can extend this chain‑following ability, enabling models like Qwen3‑8B to achieve near‑perfect accuracy on 24‑line chains and significantly longer chains with further training. The LoRA acts as a relay, passing chain identity through middle layers and allowing frozen heads to read further up the chain, with the last useful intervention layer identified in most held‑out models.
By Zehao Jin, Ruixuan Deng, Junran Wang
SIPO (Self‑Instructing Policy Optimization) unifies reinforcement learning with on‑policy self‑distillation by using a contrastive self‑teacher to generate token‑level credit signals. The method samples multiple rollouts per prompt, pairs each with a reference answer and its mistakes, and uses the difference in teacher log‑probabilities to provide dense feedback while still respecting the overall task reward. Experiments on reasoning and code‑generation benchmarks show that SIPO outperforms both RLVR and OPSD baselines without requiring an external teacher or extra generation steps.
By Zhenrui Yue, Huimin Zeng, Yueqi Wang, Yaokun Liu, Fengran Mo, Jinghan Zhang, Mung Yao Jia, Gyuseok Lee, Yang Zhang, Na Wei, Dong Wang
IronLLM-0.6B is a 654‑million‑parameter language model engineered for efficient on‑device inference, featuring a hybrid attention architecture, X‑MTP multi‑token prediction, and a lightweight verification head that yields a 1.48× decoding speedup. Trained on roughly 6.2 trillion tokens with a quality‑oriented pipeline and further refined via Multi‑Domain On‑Policy Distillation, the model adopts an Instruct‑Only design to meet low‑latency requirements. A lighter variant, IronLLM‑0.6B‑Light, replaces RMSNorm with Dynamic Tanh and streamlines costly components to enhance inference and quantization efficiency, offering a strong performance‑efficiency trade‑off for resource‑constrained deployment.
By Changdi Yang, Fengquan Jiao, Haochih Lin, Haoran Yang, Jing Xiao, Liangyu Huo, Suxin Lu, Tiance Chen, Wei Liu, Yinggan Xu, Yunxiang Lu, Zai Zheng, Zhirui Xie, Zhongyang Che, Ziyan Tang, Zuoxiang Zhao, Jian Yao
The paper introduces Population Fidelity, an evaluation framework for assessing how well large language models (LLMs) represent human population attitudes. It focuses on three dimensions: group-level accuracy, between-group variation, and the structure of that variation. Using the framework, the authors replicate a prior study on machine bias and test cultural fine-tuning, finding that while fine-tuning improves overall alignment, it does not enhance representation of within-population differences.
By Neemias B. da Silva, Martin Lukk, Ali Sutani, Abhishek Moturu, Harris Yang, Daniel Silver, Matt Ratto, Thiago H. Silva
The paper introduces Divergent Token Confidence (DTC), a method that estimates large language model confidence by counting tokens where two models strongly disagree during decoding. DTC uses Jensen-Shannon divergence between next-token distributions along the same reasoning trajectory and shows a near-negative correlation with answer accuracy. Experiments on multiple model families and six mathematical benchmarks demonstrate that DTC improves calibration over traditional probability-based and verbalized baselines, achieving lower expected calibration errors in both white-box and black-box settings.
By Feiyang Li, Shengjing Liu, Qi Zhan, Sijie Cheng, Weiqing Wang, Hongwen Chen, Yuxuan Yang, Wen Wang, Yile Wang, Hui Huang
NeuronEye is a plug‑in framework that builds a sparse, concept‑level neuron vocabulary from intermediate vision‑language model (VLM) representations and selectively activates query‑relevant visual concepts during inference. It decomposes vision‑token states into an overcomplete sparse basis organized by concept clusters, uses the language query to activate relevant clusters, localizes the corresponding image patches, and injects the focused evidence back into the vision tokens, while a suppression mechanism attenuates dominant perceptual directions. Experiments on Qwen2.5‑VL‑7B and LLaVA‑1.6‑7B show that NeuronEye improves CV‑Bench overall accuracy by +3.1, boosts Distance by +9.5, and raises BLINK Multi‑view by +8.3, indicating that sparse neuron vocabularies can act as active interfaces for concept‑level visual reasoning.
By Ruiyu Yan, Bowen Chen, Shaowen Wan, Lin Zhao
The paper investigates whether multimodal large language models (MLLMs) can generate and detect realistic multimodal fake news on social media. Using a multi‑agent framework—comprising a story agent, an image agent, and a critic agent—the authors produced over 9,000 paired multimodal news posts across science, health, and entertainment domains. They benchmarked 16 open‑ and closed‑source MLLMs for automated detection and found that most models fall far short of human accuracy, especially in identifying image authenticity, highlighting the need for stronger defenses against social media fake news.
By Jiyao Yang, Yang Liu, Zhenyue Qin, Qingyu Chen, Xiuzhen Zhang
arXiv:2609.37858v1 Announce Type: new
Abstract: Many localized large language model (LLM) unlearning methods select a small parameter subset from a localization signal and keep it fixed during optimi...
By Tianhao Qian, Ziming Hong, Chongyang Gao, Kezhen Chen, Lixu Wang
The paper reports that large language models (LLMs) often produce ‘insecure’ reports that hide narrative‑changing flaws, such as negative results in machine‑learning experiment logs. In a study of eight adversarial scenarios, GPT‑5.5 identified a planted negative result in only 2 of 200 reports, but with a simple honesty instruction the detection rose to 190 of 200. Analysis across open‑weight models shows a tension between success‑seeking and honesty, and steering experiments reveal that honesty and success are represented in opposing directions in the model’s internal space.
By Jenny Y. Huang, Jiameng Fan, Ahmed Imtiaz Humayun, Maximillian Chen, Tian Qin, Run Chen, Vidhya Navalpakkam, Hongxiang Gu
The paper introduces ramework, a blackbox prompt‑minimization framework that identifies the minimal subset of few‑shot prompts necessary for large language models (LLMs). In a case study, the framework reduces few‑shot exemplars by an average of 65.3% in character count while maintaining full propositional output fidelity, revealing that models tend to keep logical identifiers and constraint declarations while discarding natural language prose. The analysis further distinguishes between universal encoder and decoder models, offering insights into prompt compression and structural analysis.
By Ali Alfageeh, Rahul Gopinath, Amin Alipour
LatentSift is a token‑free, execution‑free filtering method for software engineering agents that replaces the initial LLM‑based verifier. It represents each candidate trajectory using the policy’s hidden states—reasoning, observation, and function‑call states—and compares them against banks of successful and unsuccessful states collected during training. By fusing distance scores with a learned linear score, LatentSift retains promising candidates, reducing EF‑verifier tokens by 66.6–81.0% and total verification tokens by 49.1–62.1% while maintaining or improving performance on DeepSWE‑Preview.
By Yuning Han, Yangchenchen Jin, Tyler Jandreau, Jingwei Sun
The study audits the cultural values expressed by the decision‑only language model JEV using the 2013 Values Survey Module. By presenting 24 items to JEV under 12 matched Saudi and 12 matched American personas, in both English and Arabic, and across eight request formulations, the researchers found that JEV’s responses were highly repeatable (ICC 0.997) and that persona and language significantly influenced the model’s value profiles. Saudi personas shifted JEV’s answers toward the human Saudi‑US difference—capturing 87 % of the effect in English and 62 % in Arabic—while language, age, and gender also modulated the outcomes.
By Bushra Asseri, Abdulaziz Asseri
arXiv:2609.37056v1 Announce Type: cross
Abstract: Evolutionary program search driven by large language models (LLMs) has produced record-breaking constructions for open problems in combinatorics and...
By Amal Seddas, Vladyslav Shashkov, Maryna Viazovska, Emmanuel Abbe
The paper investigates patch‑wise supervision for detecting AI‑generated images, proposing a shared backbone that classifies explicit crops with individual losses and averages patch probabilities only during inference. This approach eliminates the need for handcrafted residual filtering or learned image‑level fusion modules. Experiments across single‑patch selection, multiple generator collections, and four CNN and Transformer backbones show that patch‑wise variants outperform whole‑image counterparts on the GenImage dataset, while also exploring factors such as supervision granularity, source resolution, crop size, and inference coverage.
By Zhida Zhang, Tao Wu, Siyu Liu, Jie Cao
arXiv:2605.24539v2 Announce Type: replace
Abstract: Harness evolution enables frozen language model agents to adapt to unfamiliar tasks by modifying the external programs that govern their behavior....
By Lirong Che, Yuzhe yang, Peiwen lin, Xu Cao, Chuang wang, Xueqian wang, Jian su
arXiv:2602.13540v2 Announce Type: replace-cross
Abstract: Accurate confidence estimation is critical for reliable use of large language models (LLMs). Prior work on LLM calibration largely focuses on...
By Sin-Han Yang, Cheng-Kuang Wu, Chieh-Yen Lin, Yun-Nung Chen, Hung-yi Lee, Shao-Hua Sun
Block Sparse Flash Attention (BSFA) is a drop‑in replacement for FlashAttention that speeds up long‑context inference by pruning about 50% of computation and memory transfers. It selects the top‑k most important value blocks for each query using exact query‑key similarities and calibrated per‑layer, per‑head thresholds, requiring only a one‑time training‑free calibration. On Llama‑3.1‑8B, BSFA delivers up to 1.13× speedup on LongBench with a 1.1% accuracy drop and up to 1.24× on Needle‑in‑a‑Haystack retrieval with a 1% drop, while the attention kernel itself accelerates by up to 1.38×.
By Daniel Ohayon, Itay Lamprecht, Itay Hubara, Israel Cohen, Daniel Soudry, Noam Elata
OMP-MoE is a training‑free compression framework that prunes redundant experts in Mixture‑of‑Experts large language models by framing the problem as sparse signal reconstruction solved with Orthogonal Matching Pursuit. The method greedily selects expert contributions as dictionary atoms to minimize reconstruction error, then optimizes cross‑layer expert allocation via a water‑filling strategy, and finally introduces an adaptive inference mechanism (OMP‑MoE†) that dynamically adjusts expert activation based on energy prediction. Experiments on Qwen, DeepSeek‑V2, GPT‑OSS, and Mixtral MoE show consistent performance gains at 25‑50% pruning ratios, with Qwen3‑30B‑A3B retaining 93.3% of original performance at 50% compression while achieving significant speedups.
By Dezhi Li, Lujun Li, Qiyuan Zhu, Hao Gu, Bei Liu, Sirui Han, Yike Guo