Track papers across labs, topics, and trends.
Search papers, labs, and topics across Lattice.
1
0
2
Reward design in RLVR is often misinterpreted, with self-consistency elicitation dominating the signal in many scenarios, leading to significant biases in performance evaluations.