correctness

3 papers

• Credal Prediction based on Relative Likelihood
• Inference-Time Reward Hacking in Large Language Models
• Risk Management for Mitigating Benchmark Failure Modes: BenchRisk