New research is moving beyond final-outcome scoring to process-level supervision, entropy-based rewards, and evidence thresholds to catch silent failures. These methods address reliability gaps in medical decisions, multi-hop RAG, and agent trajectories.
A training framework teaches grounded QA models to answer only when evidence becomes sufficient, locating the exact point where abstention gives way to a response. On HotpotQA, 2WikiMultiHopQA and MuSiQue, with Qwen2.5-3B-Instruct and LoRA, the metho...
PRO-Step introduces step-level supervision for Retrieval-Augmented Generation. Instead of evaluating only the final answer, a generative process reward model checks logical validity and evidential grounding at every step of multi-hop reasoning. The m...
A study across 400 trajectories shows outcome-only LLM judges catch 84% of loud faults but only 45% of silent ones while flagging 33% of correct trajectories. A step-rubric judge reaches 77% silent recall with zero false alarms at 3x cost. No judge r...
ERR+ introduces a two-phase RLVR framework that rewards token-level entropy drops and relative length efficiency. Correct traces show more frequent and larger entropy drops during thinking. Tests on five datasets report improved accuracy and concisen...
Nine frontier LLMs evaluated on 2,005 oncology decision points show a collective boundary: in 42.1% of items none answers correctly, especially when choosing between guideline pathways. Two decisiveness-tuned models make unsafe commitments 3–5 times ...
