Share your thoughts, 1 month free Claude Pro on usSee more
WorkDL logo mark

Magicoder

Benchmarks

Task NameDataset NameSOTA ResultTrend
Code-Specific Instruction Tuning EvaluationMagicoder Evaluation Suite
ARC-C Accuracy54.27
48
Forgetting-aware Instruction TuningMagicoder Stability and Plasticity suites (test)
ARC-C54.27
36
Code GenerationMagicoder
Speedup5.86
12
Instruction TuningMagicoder HumanEval
Stability50.84
7
LLM RoutingMagicoder (in-distribution)
HV Score0.8514
6
Weight Poisoning AttackMagicoder
Strict ASR98.38
3
Code GenerationMagicoder
Strict ASR98.38
2
Showing 7 of 7 rows