List your company

Code Generation on BigCodeBench Instruct Full (train)

83.3Last SRAPEX-EM + Opus Judge (A5, E10)
46.48456.04265.675.158Mar 31, 2026

Evaluation results

MethodLinks
APEX-EM + Opus Judge (A5, E10)Backbone=Sonnet 4.5, Epochs=10, Variant=A5, Judge=Opus2026.0383.384
APEX-EM (EG2, E10)Backbone=Sonnet 4.5, Epochs=10, Variant=EG22026.0381.182.7
APEX-EM (E10)Backbone=GPT-4o, Epochs=102026.0381.181.5
MemRLBackbone=GPT-4o2026.0359.562.7
MemPBackbone=GPT-4o2026.0357.860.2
No Memory (A0)Backbone=Sonnet 4.5, Epochs=12026.0353.9—
Mem0Backbone=GPT-4o2026.035357.7
Self-RAGBackbone=GPT-4o2026.035055.8
No MemoryBackbone=GPT-4o2026.0348.5—
RAGBackbone=GPT-4o2026.0347.953
Pass@10Backbone=GPT-4o2026.03—57.7
ReflexionBackbone=GPT-4o2026.03—58.2