Deep Fusion Memory: A Chunk-Level Parametric Memory for Large Language Models
Open paper
Benchmark
Model
Rank
Results
arithmetic-reasoning-on-gsm8k
DFM (Qwen3-8B-Base)
–
Accuracy: 86.43
code-generation-on-humaneval
DFM (Qwen3-8B-Base)
–
Pass@1: 73.02
Rank counts only results with a code link.