iamlilAJ/active-vlm
The Perceptual Bandwidth Bottleneck in Vision-Language Models: Active Visual Reasoning via Sequential Experimental Design
I study how agents decide what matters under uncertainty.
The Perceptual Bandwidth Bottleneck in Vision-Language Models: Active Visual Reasoning via Sequential Experimental Design
ToolGate: Reducing Perceptual Tool Calls in Vision-Language Agents with Pre-Call Gating (EMNLP 2026)
A graph-based literature conflict explorer for AI paper claims.
OneManCompany talent: builds arXiv corpora, detects cross-paper conflicts, proposes new testable methods grounded in authors' own limitations. Powered by aigraph.
Stage 6 (Auto Experiment) multi-talent team: coordinator + code-writer + runner
Official Code of Memento-Research
A Principle of Targeted Intervention for Multi-Agent Reinforcement Learning
My coursework for data mining at Warwick
Config files for my GitHub profile.