Δ = B − A
CoTMode1.5B2.9B7.2B13.3B
ABΔABΔABΔABΔ
agieval5151avg@1NoCoT
+0.2%
-0.4%
+0.7%
-0.5%
aime2430avg@128CoT
—
—
—
—
—
—
—
—
aime2530avg@128CoT
—
—
—
—
—
—
—
—
aimo_progress_prize_110avg@512CoT
—
—
—
—
—
—
—
—
arc:challenge1172avg@1NoCoT
-0.7%
-1.0%
+0.7%
-0.3%
arc:easy2376avg@1NoCoT
-0.8%
-0.6%
+0.1%
-0.4%
arithmetic20000avg@1FakeCoT
-0.9%
+1.6%
+0.5%
0.0%
asdiv2305avg@1FakeCoT
+1.2%
+0.6%
+0.8%
+0.3%
bfcl_v43641avg@1FakeCoT
—
—
—
—
—
—
—
—
—
bigbench_hard13054avg@1NoCoT
—
—
—
—
—
—
—
—
—
BigCodeBench1140avg@4FakeCoT
—
—
—
—
—
—
—
—
ceval_zho_mcf1343avg@1NoCoT
+0.6%
-0.2%
+0.5%
-0.6%
commonsenseqa1221avg@1NoCoT
-0.2%
-0.5%
+0.7%
0.0%
gaia165avg@1FakeCoT
—
—
—
—
—
—
—
—
gpqa:diamond198avg@1CoT
+2.0%
-1.6%
+0.4%
+2.1%
gpqa:main448avg@1CoT
+0.7%
-0.7%
+0.9%
-1.6%
gsm8k1319avg@1CoT
+1.1%
+1.8%
+0.6%
-2.3%
gsm_plus9233avg@1CoT
-0.3%
+0.4%
+0.4%
-1.7%
hellaswag10042avg@1NoCoT
-1.9%
+0.1%
+0.3%
-0.2%
HumanEval164avg@16FakeCoT
-2.1%
-1.8%
+0.1%
+0.3%
HumanEval164avg@16CoT
-10.4%
-6.1%
-10.8%
-8.8%
HumanEval+164avg@16FakeCoT
-1.3%
-1.4%
+0.8%
+1.9%
HumanEval+164avg@16CoT
-10.1%
-5.2%
—
—
-9.1%
ifbench_multiturn1387avg@1CoT
+3.4%
+1.9%
-2.7%
+2.9%
ifbench_test300avg@1CoT
+3.0%
+3.7%
+3.0%
+3.0%
ifeval541avg@1CoT
+7.2%
+0.4%
-0.4%
+3.9%
lcb:codegeneration268avg@1CoT
0.0%
+4.9%
-0.4%
0.0%
math5000avg@4CoT
+2.5%
+0.8%
+1.0%
-0.4%
math_500500avg@8CoT
—
—
—
—
—
—
—
—
mathqa2985avg@1NoCoT
-0.3%
-0.3%
+0.7%
-0.3%
MBPP500avg@8FakeCoT
+5.0%
-2.7%
-1.7%
-1.4%
MBPP500avg@8CoT
-3.6%
—
—
—
—
—
—
MBPP+378avg@8FakeCoT
+5.2%
+2.1%
-3.0%
-0.5%
MBPP+378avg@8CoT
—
—
—
—
—
—
—
—
mc_taco9442avg@1CoT
—
—
—
—
—
—
—
—
med_mcqa4183avg@1NoCoT
-1.4%
+0.5%
0.0%
+0.3%
med_qa2545avg@1NoCoT
-0.4%
+0.2%
+0.2%
+0.4%
mmlu14042avg@1NoCoT
+0.8%
+0.2%
+0.7%
+0.2%
mmlu_pro12032avg@1NoCoT
+1.7%
+2.5%
-2.7%
-0.1%
mmlu_redux_25700avg@1NoCoT
-0.2%
+0.4%
+0.4%
-0.3%
MultiPL-E HumanEval2864avg@1FakeCoT
—
—
—
—
—
—
—
—
—
—
mutual886avg@4CoT
—
—
—
—
—
—
—
—
olympiad_bench2673avg@1CoT
-0.5%
+0.4%
+0.5%
-0.6%
openbookqa1000avg@1NoCoT
-0.2%
0.0%
+0.6%
-0.3%
race1045avg@4CoT
—
—
—
—
—
—
—
—
tau2_bench278avg@1FakeCoT
—
—
—
—
—
—
—
—
tau3_bench375avg@1FakeCoT
—
—
—
—
—
—
—
—
tau_bench165avg@1FakeCoT
—
—
—
—
—
—
—
—
truthfulqa:mc817avg@1NoCoT
-0.1%
-0.5%
+0.4%
-0.5%
winogender960avg@2CoT
—
—
—
—
—
—
—
—
winogrande1267avg@1NoCoT
+0.4%
-1.0%
-0.1%
+0.6%
wmdp3668avg@1CoT
—
—
—
—
—
—
—
—