Search papers, labs, and topics across Lattice.
Politecnico di Milano
2
0
5
MLLMs in Medical VQA can be fine-tuned to reduce overconfidence by over 60%, fundamentally changing how we assess model reliability in clinical settings.
Stripping away the complexity of GRPO reveals that simple REINFORCE with group relative advantage can actually *improve* LLM reasoning, challenging the assumption that sophisticated loss functions are always better.
