Tag: Calibration-Aware Policy Optimization
Browse our exclusive articles!
Odysseus: Scaling VLMs for 100+ Turn Game Decisions
Discover Odysseus, a new RL framework that scales vision-language models for long-horizon decision-making in games like Super Mario Land.
RHyVE: Reliable Verification & Deployment of LLM Rewards
Discover RHyVE, a protocol for competence-aware verification and phase-aware deployment of LLM-generated reward hypotheses in reinforcement learning.
TLPO: Boosting Language Consistency in Large Language Models
Discover how Token-Level Policy Optimization (TLPO) reduces language confusion in large language models, improving multilingual accuracy and performance.
Ranking-Based Explanation Quality Assessment with Listwise Rewards
Discover a novel ranking approach to explanation quality assessment using listwise rewards, improving AI model evaluation and policy optimization.
V-tableR1: Advanced Multimodal Table Reasoning AI
Discover V-tableR1, a novel AI framework enhancing multimodal table reasoning with critic-guided policy optimization for superior visual and logical infere...