Token-level hallucination detectors trained with the TokenHD pipeline. Models range from 0.6B to 8B parameters, all based on Qwen3 backbones.
Rui Min PRO
mr233
AI & ML interests
None yet
Recent Activity
upvoted a paper about 2 months ago
GeoBrowse: A Geolocation Benchmark for Agentic Tool Use with Expert-Annotated Reasoning Traces updated a dataset 3 months ago
mr233/TokenHD-training-data