8 papers
DIVE: Dynamic Iterative Visual Evidence Construction for Efficient Vision-Language Models
Chen Zhong, Xiao An, Zijie Wang +3
Visual inputs in vision-language models (VLMs) are often encoded into substantially longer token sequences than text, making visual tokens a major bottleneck for efficient inferenc…
Convolution for Large Language Models
Yuchuan Tian, Yingte Shu, Wei He +7
Large language models (LLMs) largely rely on Transformers, where self-attention provides global token interaction but does not explicitly encode the locality of natural language. W…
Agentic Routing: The Harness-Native Data Flywheel
Xinchen Liu, Hang Zhou, Yingjie Zong +12
The paper introduces a step‑level routing framework for large language model agents that selects the most suitable model(s) based on the full execution harness state, using logged…
Claw-SWE-Bench: A Benchmark for Evaluating OpenClaw-style Agent Harnesses on Coding Tasks
Mengyu Zheng, Kai Han, Boxun Li +13
General-purpose agents such as OpenClaw are increasingly used as autonomous tool users, but their coding ability is difficult to measure under SWE-bench: a generic agent does not b…
Self-Consistent Latent Reasoning: Long Latent Sequence Reasoning for Vision-Language Model
Chenfeng Wang, Wei He, Xuhan Zhu +10
In language reasoning, longer chains of thought consistently yield better performance, which naturally suggests that visual latent reasoning may likewise benefit from longer latent…
SenseBench: A Benchmark for Remote Sensing Low-Level Visual Perception and Description in Large Vision-Language Models
Chen Zhong, Xiao An, Jiaxing Sun +3
Low-level visual perception underpins reliable remote sensing (RS) image analysis, yet current image quality assessment (IQA) methods output uninterpretable scalar scores rather th…