ReMix: Reinforcement routing for mixtures of LoRAs in LLM finetuning
Open paper
Benchmark
Model
Rank
Results
arithmetic-reasoning-on-gsm8k
ReMix (Llama 3 8B)
–
Accuracy: 65.66
code-generation-on-humaneval
ReMix (Llama 3 8B)
–
Pass@1: 32.93
Rank counts only results with a code link.