Thermo-FL is a thermal‑aware federated LoRA fine‑tuning framework that adapts local adapter training and sparse update transmission based on device temperature. It introduces TERRA, a robust aggregation pipeline that uses norm filtering, mask‑aware directional validation, adaptive clipping, and mask‑aware aggregation to defend against Byzantine and communication‑layer adversaries. Experiments on a large‑scale emulator and a Jetson testbed show Thermo-FL improves robustness under adversarial sparse aggregation, stabilizes device temperature, reduces upload size, and preserves utility on GSM8K and BoolQ tasks.
By Shiva Shrestha, Kazi Shaharair Sharif, Zongxing Xie, Jiajing Huang, Anhao Xiang, Honghui Xu
The paper introduces Conformalized Agentic Search (CAS), a framework that applies Conformal Prediction to improve the reliability of search agents during reinforcement learning fine-tuning. CAS uses an Adaptive Prediction Set (APS) to dynamically truncate retrieved documents based on statistical coverage, and Adaptive Conformal Inference (ACI) to construct confidence-aware prediction sets that penalize low‑confidence trajectories in the Group Relative Policy Optimization objective. Experiments on single‑hop and multi‑hop QA datasets show that CAS enhances reasoning accuracy and reduces redundant tool invocations, offering a more reliable and efficient agent paradigm.
By Zixi Zhu, Jiayuan Su, Jian Zhang, Yu Lin, Hongwei Wang
arXiv:2606.15765v2 Announce Type: replace
Abstract: Vision foundation models (VFMs) exhibit complementary strengths shaped by their pretraining objectives. Yet prevailing methods for multi-task dense...
By Donghyun Han, Yuseok Bae, Jung Uk Kim, Hyung-Il Kim
arXiv:2606.13971v3 Announce Type: replace
Abstract: While personalizing Image-to-Video (I2V) diffusion models with specific visual effects is increasingly demanded for high-end generation, current pr...
By Xiaomeng Yang, Yanyu Li, Gordon Guocheng Qian, Ivan Skorokhodov, Viacheslav Ivanov, Avalon Vinella, Xuan Zhang, Yanzhi Wang, Sergey Tulyakov, Anil Kag
The paper reports the winning solution to the MoCha 2026 Parkinsonian Gait Benchmark, achieving a macro‑F1 score of 0.6945 on unseen clinical sites. The approach relies on a frozen public motion encoder followed by a single 4×512 linear layer, and gains are largely attributed to three key steps: exact replication of the benchmark’s head recipe, averaging per‑walk posteriors at the subject level, and a label‑free transductive calibration of feature means and decision thresholds. Extensive ablation studies show that fine‑tuning the encoder or using alternative encoders does not improve performance, and the subject‑level aggregation is identified as the primary contributor to the top score.
By Junlong Shen
Re$^3$Cap introduces a retrieval‑guided refinement strategy for image captioning that leverages multi‑modal retrieval as a reasoning signal. The method, built on a Caption Refinement Suggester and a Caption Quality Assessor, detects hallucinations and omissions to produce more accurate and detailed captions without extra annotations. Experiments show it surpasses supervised fine‑tuning and improves relation reasoning by 8.64% on the COCO‑LN500 benchmark.
By Haonan Jia, Shichao Dong, Zenghui Sun, Jiawen Zheng, Ziqi Miao, Gege Shi, Qiuyu Zhao, Jinsong Lan, Xiaoyong Zhu, Bo Zheng
arXiv:2608.21300v1 Announce Type: new
Abstract: Foundation models for medical image segmentation, like prompt-based MedSAM, generalize well across domains and modalities, often in zero or few-shot se...
By Marko Haralovi\'c, Sounic Akkaraju, Carlo Baretta, Vasil Zapryanov, Alexia Briassouli
arXiv:2608.20974v1 Announce Type: cross
Abstract: Video Joint Embedding Predictive Architecture (V-JEPA) learns powerful spatiotemporal representations from video through self-supervised latent featu...
By Xinlin Wang, Yujiao Xiang, Yuheng Zhou, Jingqi Wang, Minqing Huang, Jiajie Huang, Dongxu Wei, Tingguang Zhou, Xiyang Wang, Gong Chen, Zhi Xu, Feiyang Tan, Hangning Zhou, Mu Yang
arXiv:2605.01913v2 Announce Type: replace-cross
Abstract: Fine-tuning safety-aligned language models for downstream tasks often leads to substantial degradation of refusal behavior, making models vul...
By Sadia Asif, Mohammad Mohammadi Amiri
ForeTime‑VLA is a causal vision‑language‑action policy that distills future‑aware representations from a frozen Fast‑WAM teacher, enabling it to anticipate contact events during conveyor‑belt manipulation. The method compresses current and future video latents into a 64‑dimensional target, uses an eight‑frame history encoder to predict this target along with manipulation phase and time‑to‑transition, and conditions a VLM prefix on future tokens and phase. On a deduplicated conveyor‑belt dataset, ForeTime‑VLA reduces test MAE by 2.63% and L2 by 3.02%, while real‑robot experiments show significantly higher grasp success rates compared to the next‑best reference.
whyItMatters":"The approach demonstrates that distilling future‑token knowledge from a world‑action model can improve dynamic manipulation performance without the computational cost of running the teacher at inference time."
By Siyuan Ma, Yutian Zhang, Boshi Zhang, Qinglian Wu, Jiaqi Zhai, Dong Wei, Xiaojin Huang
The paper investigates how agentic systems decide between acting and abstaining, focusing on the fidelity of their reasoning explanations. Using Qwen3‑8B in a multi‑party conversation setting, the authors compare direct decision policies, reasoning policies, supervised fine‑tuning, and reinforcement learning, finding a trade‑off: strong direct policies yield higher performance but no traceable reasoning, while reasoning policies provide an audit trail at the cost of lower recall. The study also uncovers that exposing reasoning can alter the agent’s policy and that common faithfulness metrics may overstate the alignment between reasoning and decisions.
By Shreya Mendi, Brinnae Bent
arXiv:2608.21024v1 Announce Type: new
Abstract: Modern neural network training increasingly uses matrix-aware optimizers, yet their conditioned matrix step is typically added directly to the weight,...
By Guoxiang Xu, Bince Qu, Qi Sun, Cheng Zhuo
CLEAR is a conditional safety adaptation framework that employs a lightweight hidden‑state gate to continuously control the activation strength of a safety low‑rank adapter. It aims to reduce harmful completions while preserving the performance of the frozen backbone on benign prompts. Experiments on safety and utility benchmarks, including HarmBench and GSM8K, show that CLEAR significantly lowers HarmBench ASR and improves utility compared to globally applied safety tuning methods such as SFT or standard LoRA.
By Chengxiao Wang, Enyi Jiang, Xiaojing Liao, Sanmi Koyejo
arXiv:2608.21325v1 Announce Type: new
Abstract: Users increasingly turn to large language models for emotional support, yet little is known about how these models actually conduct a psychotherapy int...
By Afonso Baldo, Hugo Pitorro, Areti Vassilopoulos, Anabela C. Areias, Maya D'Eon, Fab\'iola Costa, Ricardo Rei, Nuno M. Guerreiro
arXiv:2607.09885v3 Announce Type: replace
Abstract: We present Index-1.9B, a series of open small language models developed at Bilibili. The series comprises four models: Index-1.9B-Base, a foundatio...
By Tianjiao Li, Lusheng Zhang, Shien He, Xiaojing Liu, Tianxing Yan, Mengran Yu, Ziang Cui, Kai Zhao, Xipeng Wang, Yang Liu, Yuxin Li
UpgradeBench is a decision‑centric longitudinal benchmark that evaluates how fine‑tuned language‑model specialists should be handled when new base‑model releases occur. It covers four consecutive Qwen releases, a continuation checkpoint, six tasks, two model sizes, and OLMo checkpoints with known training lineage, and examines whether retraining, adapter transfer, or other recovery strategies improve specialist performance. The benchmark reveals that upgrade gains vary by task and release interval, that direct adapter copying is sensitive to pretraining distance, and that teacher relabeling can recover specialists without new annotations.
"whyItMatters":"The study provides actionable insights into the cost‑effective management of specialist models across model releases, showing how to balance retraining effort with performance gains."
By Ye Chen, Weining Zhang
The paper investigates whether distilling reasoning traces from large teacher models is worth the extra compute compared to standard instruction fine‑tuning (IFT). By generating paired IFT and reasoning outputs from the same teacher and training student models at five scales, the authors find that, at matched FLOPs, IFT generally lies on or near the Pareto frontier across most configurations. Reasoning traces only reach the frontier on open‑ended tasks for models 7B and larger, and a curriculum mixing 25–50% reasoning data with IFT can capture most of the accuracy benefit at a lower compute cost.
By Nicolas Boizard, Hippolyte Gisserot-Boukhlef, Kevin El Haddad, C\'eline Hudelot, Pierre Colombo
MatMMExtract is an open‑source pipeline that disassembles compound scientific figures into individual sub‑panels and generates structured, grounded image‑text pairs using a large language model guided by a materials science taxonomy. Applied to 14,810 open‑access articles, it produced 391,606 panel‑level pairs with sub‑captions, a two‑level visualisation category (19 classes, 100+ subtypes), and scientific summaries. The project also introduces MaterialScope, a 2,811‑figure detection dataset, and demonstrates that Gemini 3.1 Flash Lite yields high‑quality annotations with low hallucination, while a dual‑encoder baseline outperforms zero‑shot CLIP on the resulting MatSciFig dataset.
By Subham Ghosh, Shubham Tiwari, Mohammad Ibrahim, Abhishek Tewari
arXiv:2508.17488v4 Announce Type: replace
Abstract: This paper tackles the critical challenge of optimizing multi-modality trackers by effectively adapting pre-trained models for RGB data. Existing f...
By Zhiwen Chen, Jinjian Wu, Zhiyu Zhu, Yifan Zhang, Guangming Shi, Junhui Hou
arXiv:2608.20699v1 Announce Type: new
Abstract: Animating articulated 3D meshes via text requires satisfying strict kinematic constraints, modeling causal interactions between parts, and achieving in...
By Chunyu Zou, Peng Dai, Yi-Hua Huang, Ze Yuan, Jingwei Huang, Yeming Yao, Xiaojuan Qi