Share your thoughts, 1 month free Claude Pro on us
See more
Home
/
Benchmarks
Knowledge on MMLU-Pro (test)
Loading...
58.6
Accuracy
GAC + Token-φ
11.7896
23.9423
36.095
48.2477
Jun 13, 2025
Aug 9, 2025
Oct 6, 2025
Dec 3, 2025
Jan 29, 2026
Mar 28, 2026
May 25, 2026
Accuracy
Updated 7d ago
Evaluation Results
Method
Method
Links
Accuracy
GAC + Token-φ
Variant=Token-φ
2026.05
58.6
GAC w/o φ
Variant=w/o φ
2026.05
57.8
HPT
Category=Recent hybrid...
2026.05
56.4
CHORD
Category=SFT–RL mixing...
2026.05
56.2
LUFFY
Category=Recent hybrid...
2026.05
56
KL-ctrl
Category=Rule-based co...
2026.05
55.8
SRFT
Category=Recent hybrid...
2026.05
55.6
Nash-MTL
Category=Multi-objecti...
2026.05
55.4
GradNorm-ctrl
Category=Rule-based co...
2026.05
55.3
CAGrad
Category=Multi-objecti...
2026.05
55.1
SFT-best + RL
Category=SFT–RL mixing...
2026.05
51.3
GRPO (pure RL)
Category=SFT–RL mixing...
2026.05
45.8
DPO
Category=RL-free align...
2026.05
42.1
IPO
Category=RL-free align...
2026.05
41.5
SFT-best
2026.05
38.4
Qwen2.5-7B-Inst.
2026.05
24.7
Dense baseline
Model Type=Dense, Comp...
2025.06
14.12
MoE w/ optimal AR
Model Type=MoE, Activa...
2025.06
13.59
MoE w/ optimal AR
Model Type=MoE, Activa...
2025.06
13.59
Feedback
Search any
task
Search any
task