QVal: Evaluating Dense Supervision Signals for Long-Horizon LLM Agents
cs.LG, cs.AI, cs.CL
Submitted: 2026-06-30
Updated: 2026-11-13
Terminology
Sources
- Vision-Language Models as a Source of Rewards
- OpenGVL -- Benchmarking Visual Temporal Progress for Data Curation
- DeepSeek-V3.2: Pushing the Frontier of Open Large Language Models
- Reinforcement Learning via Self-Distillation
- AI2-THOR: An Interactive 3D Environment for Visual AI
- VLM Judges Can Rank but Cannot Score: Task-Dependent Uncertainty in Multimodal Evaluation
- RewardBench: Evaluating Reward Models for Language Modeling
- Auto MC-Reward: Automated Dense Reward Design with Large Language Models for Minecraft
- VL-RewardBench: A Challenging Benchmark for Vision-Language Generative Reward Models
- CriticBench: Benchmarking LLMs for Critique-Correct Reasoning
- RM-Bench: Benchmarking Reward Models of Language Models with Subtlety and Style
- AgentRewardBench: Evaluating Automatic Evaluations of Web Agent Trajectories
- VIP: Towards Universal Visual Reward and Representation via Value-Implicit Pre-Training
- RewardBench 2: Advancing Reward Model Evaluation
- Revisiting the Learning Objectives of Vision-Language Reward Models
- DeepSeekMath: Pushing the Limits of Mathematical Reasoning in Open Language Models
- Self-Distillation Enables Continual Learning
- When To Solve, When To Verify: Compute-Optimal Problem Solving and Generative Verification for LLM Reasoning
- Expanding the Capabilities of Reinforcement Learning via Text Feedback
- TIPS: Turn-Level Information-Potential Reward Shaping for Search-Augmented LLMs
Related papers
- A Physics-Regulated Neural Framework for Learning 3D Grain Growth Dynamics
- FlexMoE: Learning One-for-All Multi-Budget Expert Pruning for Flexible MoE Deployment
- OptiMind: Improving Small Language Models for Optimization Modeling through Formulation Guidance and Multi-Turn Learning
- Staying in the Sweet Spot: Boost Reasoning Evolution via Responsive Rollout Difficulty Manipulation
- Examining the Impact of Data and Training Methods on Emergent and Subliminal Misalignment
- Rethinking Black-Box On-Policy Distillation: A Rubric-based Analysis