Meta-Reward
A signal used to evaluate the quality of a reasoning process rather than the correctness score of the final answer. It is used to decide meta-actions such as revision, backtracking, and stopping, and can provide dense feedback at intermediate steps.
Timeline (last 90 days)
GraphEvoResearcher designs a graph-aware meta-reward function that combines five types of signals to decide the next action: correctness, efficiency, reflection quality, diversity, and structural signals (connectivity, acyclicity, branch coverage, and path convergence).
Meta-reward here does not refer to the correctness score of the task itself but to an evaluation signal of "reasoning process quality," used to decide the next meta-action such as revise, backtrack, or stop.
All Facts (2)
GraphEvoResearcher designs a graph-aware meta-reward function that combines five types of signals to decide the next action: correctness, efficiency, reflection quality, diversity, and structural signals (connectivity, acyclicity, branch coverage, and path convergence).
50%UnverifiedMeta-reward here does not refer to the correctness score of the task itself but to an evaluation signal of "reasoning process quality," used to decide the next meta-action such as revise, backtrack, or stop.
50%