Vela: User Behavior as Supervision Sparks LLM Intelligence
Open paper
Benchmark
Model
Rank
Results
mathematical-reasoning-on-aime-2026
Vela-35B-A3B (extra data)
#17
Accuracy: 92.6
instruction-following-on-ifbench
Vela-35B-A3B (extra data)
#19
Accuracy: 69.2
Rank counts only results with a code link.