Dimensions overall score 8.0

GitHub Code Pulse

No public code linked for this paper yet.

• MR-Search introduces an in-context meta reinforcement learning formulation for agentic search with self-reflection.method 95%
• MR-Search trains a policy that conditions on past episodes and adapts its search strategy across episodes, rather than optimizing within a single independent episode.method 90%
• MR-Search performs cross-episode exploration by generating explicit self-reflections after each episode and using them as context for subsequent attempts.method 90%
• MR-Search introduces a multi-turn RL algorithm that estimates a dense relative advantage at the turn level for fine-grained credit assignment.technical 90%
• Empirical results show MR-Search achieves relative improvements of 9.2% to 19.3% across eight benchmarks compared to baseline RL methods.result 95%
• MR-Search demonstrates advantages over baseline RL methods in various benchmarks.result 85%
• The self-reflection mechanism in MR-Search allows search agents to improve in-context exploration at test-time.result 85%

Competitive landscape

Competitor map is still being generated for this paper. Enable generation or check back soon.

Startup potential card