Trustworthy Machine Learning and Reasoning (TMLR) Group, an online-offline-mixed machine learning research group, locates in different cities, including Hong Kong, Melbourne, Shanghai, Nottingham and Sydney. We share the vision for the future ML technology: building trustworthy learning and reasoning algorithms, theories and systems.
Pinned Loading
-
[ICML 2026] "The Easy, the Hard, and the Learnable: Confidence and Difficulty-Adaptive Policy Optimization for LLM Reasoning"
Python 3
-
[ICML 2026] "Concept Concentration for Faithful Representation Intervention"
Python 1
-
Forked from resistzzz/Co-rewarding
[ICLR 2026] "Co-rewarding: Stable Self-supervised RL for Eliciting Reasoning in Large Language Models"
-
Forked from junnie00/ConV
[NeurIPS 2025 Spotlight] "Detecting Generated Images by Fitting Natural Image Distributions"
Python 13
-
[NeurIPS 2025] "Generative Model Inversion Through the Lens of the Manifold Hypothesis"
Python 5
Repositories
Showing 10 of 128 repositories
-
AlphaDiana Public
AlphaDiana: A System for Evaluating Agentic Reasoning
-
RewardFlow Public
[arXiv:2603.18859] "RewardFlow: Topology-Aware Reward Propagation on State Graphs for Agentic RL with Large Language Models"
-
AgentHijack Public Forked from super-jw/AgentHijack
[ICML 2026] "AgentHijack: Benchmarking Computer Use Agent Robustness to Common Environment Corruptions"