The paper presents a taxonomy-driven framework for identifying, categorizing, and explaining bias in AI-generated Python code. By extending an existing dataset and manually annotating bias categories and justifications, the authors evaluate both proprietary and open-source large language models (LLMs) for automated bias detection and explanation. Results show that models such as Gemini and Qwen3-coder achieve high classification accuracy and produce justification and code identification similarities that closely match human-authored reasoning.
By Manaal Basha, Aimee M. Ribeiro, Gema Rodriguez-Perez
VLALight is a lightweight end‑to‑end vision‑language‑action framework designed for traffic signal control. It fuses multiple camera views and textual instructions to directly predict signal actions, avoiding intermediate image‑to‑text conversions. The model, with only 0.5 B parameters, achieves superior emergency vehicle service, cutting pooled waiting time by 21.1% compared to cascaded methods while running in real time on local hardware.
By Kemou Jiang, Maonan Wang, Xingchen Zou, Jiayue Zhu, Yuhang Fu, Sicheng Wang, Xi Chen, Yirong Chen, Zhiyong Cui
The study investigates how Google’s Gemma 4‑e4b language model resolves conflicts between two documents. Using a counterbalanced design, researchers found that the semantic framing of a source (e.g., labeling it as an official guideline) dominates over the order in which documents appear. While the model shows a primacy bias toward the first document, this bias varies widely with wording and is amplified only when the documents are structurally identical.
By Amanda Fitch
SustainAI is a water‑aware, closed‑loop framework that embeds environmental accountability into AI deployment. It combines real‑time water metering, a hallucination‑aware penalty model, and a water‑aware routing algorithm that considers regional water stress. In tests with small language models, water footprints varied 11‑fold across data centers, and 1,335 inference runs consumed about 399 mL of water but yielded only 240 correct outputs, highlighting the resource cost of inaccurate responses.
By Farnaz Farid, Tashfia Towkee, Sania Nasreen, Sami bin Azad
The paper introduces LIRSeg, a method that replaces explicit Chain-of-Thought reasoning in multimodal large language models with a compact set of learnable latent tokens for reasoning segmentation. LIRSeg is trained in two stages—spatial alignment and GRPO—while employing extreme-advantage sampling, decoupled exploration-stability updates, and latent diversity amplification to enhance token informativeness. Experiments show that LIRSeg improves segmentation accuracy and reasoning efficiency, achieving significant gIoU gains over the VisionReasoner baseline and reducing reasoning tokens by about 16×.
By Tianhang Guo, Yulin He, Wei Chen, Wenjuan Zhou, Yuhang Li, Xinbiao Gan
AI agents built on shared models tend to follow a common forecast, causing them to cluster on one road while avoiding a less crowded alternative. In a two-road congestion game, adding a warning that others might follow a routing tip led 50 GPT agents to crowd one road, raising average travel time from 64 to 95 minutes. Human participants largely maintained balance, but in mixed groups the imbalance grew as more agents were involved, with agents incurring higher travel times than humans.
By Takahiro Ezaki, Naoto Imura, Katsuhiro Nishinari
UltraG-Bench is a large‑scale, multi‑task benchmark designed to evaluate pixel‑level evidence grounding in ultrasound images. It comprises 40 public segmentation datasets covering 13 anatomical categories and includes three progressive tasks—instruction‑guided segmentation, evidence‑grounded VQA, and evidence‑grounded report generation—with a total of 736,726 annotations. Evaluation of 14 state‑of‑the‑art models shows a significant gap between semantic understanding and fine‑grained pixel‑level localization, and the authors propose UltraG‑Agent, which combines a vision‑language model with the ultrasound‑specific segmentation model UltraSAM to improve both semantic prediction and visual grounding.
By Quanhao Zhu, Bo Xu, Rui Lin, Chenyuan Wang, Yu Shao, Boling Zhu, Jiuyan Sun, Liang Zhao, Hongfei Lin, Feng Xia
The paper introduces EoupCT, a framework that estimates and orthogonalizes unknown pre‑training gradients to mitigate catastrophic forgetting during continual fine‑tuning of large language models. It generates pseudo data most susceptible to forgetting using a learnable soft prompt with Gumbel‑Softmax, then jointly optimizes model parameters and the prompt via a first‑order Pareto optimizer to enforce orthogonality between new task updates and the estimated gradients. Experiments on multiple LLMs show that EoupCT preserves both task‑specific performance and the models’ inherent general‑purpose knowledge.
By Bing Wang, Changchun Li, Xin-Qiang Cai, Lin Yuanbo Wu, Ximing Li, Gang Niu, Masashi Sugiyama
MVVBench is a new benchmark for multi‑view video reasoning that tests vision‑language models on tasks requiring integration of spatial and temporal evidence across multiple, often non‑overlapping camera streams. The benchmark contains questions that cannot be answered from any single view or single moment, forcing models to jointly reason across views and time. It evaluates six capabilities—including attribute identification, relative distance, camera pose, and compositional counting—and provides human‑authored QA, rigorous verification, and detailed error analysis.
"whyItMatters":"The benchmark offers a rigorous evaluation of 4D multi‑view reasoning and a foundation for future progress toward reliable embodied perception."
By Hyungjin Chung, Byeongjun Park, Joonseok Lee, Hojun Kim, Jaeho Choi, Byung-Hoon Kim
The paper proposes a theoretical framework called the Linear Representation Hypothesis (LRH) for vision‑language‑action (VLA) models, extending the concept from large language models to systems where physical quantities of interest (QoI) evolve with the dynamics. It introduces a signature‑based formulation that unifies representations and policies, proving that future QoI evolution can be linearly probed from representations and that a generalized linear model for stochastic action chunks allows monotonic steering of QoI. The authors validate their theory with an explicit oracle representation in a planar control‑affine navigation experiment, demonstrating the predicted linear probing and steering mechanisms.
By Minseok Jeong, Hyewon Choi, Hiroyasu Tsukamoto, SooJean Han
G$^2$PTQ is a post‑training quantization framework that improves large language models by combining first‑ and second‑order information in a globally supervised, block‑wise optimization. It refreshes gradient and Hessian estimates before each Transformer block and uses a trust‑region scaling mechanism to stabilize gradient steps, preventing exploding weight updates. The method achieves better alignment with full‑precision models and outperforms state‑of‑the‑art baselines across various model families and bit‑widths.
By Ruikang Liu, Haoli Bai, Yuxuan Sun, Qian Zhang, Wenzheng Cai, Yanqi Hao, Feiyu Wang, Weidong Zhong, Zhuang Wang, Tong Yang, Xiangsheng Zhou
DynBranch introduces speculative subgraph reuse for dynamic agentic LLM serving by making unresolved branches addressable before they resolve. This allows candidate subgraphs to run during branch resolution, enabling reuse of completed subgraph results across subsequent requests. Experiments with Qwen3-32B on 4x H200 GPUs show up to 32% mean latency reduction over the best prior system and 46–66% over a no‑reuse baseline, while maintaining correct workflow outputs.
By Junyi Shen, Noppanat Wadlom, Zhengyuan Su, Yao Lu
DepthEvidence is a 4B multimodal language model that integrates dense metric depth predictions into language generation. It employs a camera‑conditioned decoder to produce full‑resolution depth maps and a dense‑to‑language interface that converts these predictions into object‑aligned geometry tokens. The model is trained with geometric supervision and instruction tuning, and it sets new state‑of‑the‑art results on a Depth‑VQA benchmark and on instance‑level metric depth estimation across nine datasets.
By Jiangning Wei, Yuan Yao, Miaomiao Cui, Mingsheng Li, Humen Zhong, Shuai Bai, Zhibo Yang
Pocket-STVG (P-STVG) is a lightweight cascade architecture for Spatio-Temporal Video Grounding that combines efficient pre‑trained components: a temporal‑aware video encoder based on MobileViCLIP, a spatial encoder‑decoder from MDETR, and a shared aligned text encoder. Temporal localization is achieved with a lightweight 1D U‑Net or a simple thresholding strategy, allowing the model to work in both weakly supervised and zero‑shot settings. With fewer than 90 M parameters, P-STVG matches or surpasses prior weakly supervised and zero‑shot methods while offering a more memory‑ and compute‑efficient pipeline for large‑scale video collections.
By Alberto Presta, Michal Byra, Grzegorz Stefa\'nski, Karol Szurkowski, Eryk Ko{\l}odziejczyk, Krzysztof Arendt
The paper investigates how to choose the best quantized model from a family of compressed versions when target labels are scarce or unavailable. It finds that a simple rule based on minimum teacher distortion consistently selects the same eight‑bit, per‑channel, unclipped configuration, though this does not minimize empirical target cross‑entropy. The study also shows that confidence‑based estimators perform poorly in overconfident regimes, while output‑distribution estimators can outperform the teacher in some architectures, and that combining distortion with a supervised term can improve selection. Across 134 candidate families, teacher‑anchored selection reduces mean regret with very few labels, though the benefit diminishes after about 25 labels.
By Alejandro Rodriguez Dominguez, Muhammad Shahzad, Xia Hong
AgentRecommender is a method that uses large language model (LLM) agents to create customizable recommender systems on the user side, eliminating the need for additional data. By leveraging the agents' investigative capabilities and internal knowledge, users can build recommender systems tailored to their own preferences without relying on platform-driven algorithms. This approach addresses issues such as clickbait, filter bubbles, and the spread of fake news that arise from platform-centric recommender systems.
By Ryoma Sato
The paper introduces acoustic-to-text KV compression for full‑duplex speech models, converting acoustic key‑value states into compact textual memory during listening‑time slack. When the KV cache exceeds a target budget, older acoustic states are evicted while transcripts and recent acoustic context are retained. Experiments on ten‑minute LongSpeech sessions show a 64.6% reduction in peak streaming KV‑cache size and improved transcription, temporal question answering, and summarization, with comparable pause‑handling, turn‑taking, and interruption performance in Full‑Duplex‑Bench.
By Yejin Lee, Seungbeom Kim, Yongha Lee, Kyuhong Shim
The paper introduces a post‑training softmax reparameterization technique that selects a functionally equivalent output head before quantization. By subtracting a scalar multiple of the vocabulary‑row mean from each output row and tuning this coefficient via validation KL, the method preserves the full‑precision softmax distribution while enabling efficient W4 quantization. Experiments on seven heads show significant error reductions and latency improvements, with the approach remaining complementary to other quantization strategies and transferable across datasets.
By Asim Kadav, Christian Flores, Chirag Arora, Varun Kotte, Hongbo Zheng, Lan Yan, Priya Shanmugasundaram, Tracy Holloway King
EffectMatch is a runtime system that monitors and validates persistent changes made by large language model agents during software interactions. It collects all side‑effects within a controlled execution boundary and compares them against the application’s approved state, deciding whether to commit the changes and allow subsequent steps. In tests on 206 public business tasks, EffectMatch preserved correct executions and prevented all incorrect commits, with ablation studies showing the importance of each component.
By Haoran Zhang, Hengtong Zhang, Zhiyu Liang, Yu Yan, Decheng Zuo, Hongzhi Wang
AgentXploit is a two‑role auditing system that separates repository‑level attack‑path discovery from runtime exploitation for AI agents. The Analyzer Agent traces attacker‑controlled inputs to sensitive operations and records candidate attack paths, while the Exploiter Agent turns these paths into concrete attacks and refines them using runtime feedback. The system is evaluated on AgentXploit‑Bench, a benchmark of 72 reproducible vulnerabilities across 12 open‑source AI‑agent systems, achieving 59.3% end‑to‑end success compared to 38.4% for Codex, and 79.2% attack success on AgentDojo versus 52.7% for AgentVigil.
By Weida Liang, Shi Qiu, Zhun Wang, Simon Sure, Xiaoyuan Liu, Tianneng Shi, Zhaorun Chen, Wenbo Guo, Dawn Song