AA-LCR厂商自报 · 越高越好
52%
NeMo Gym / NeMo Evaluator; BF16
成绩出处BrowseComp厂商自报 · 越高越好
36.97%
NeMo Gym / NeMo Evaluator; BF16
成绩出处GPQA厂商自报 · 越高越好
75.44%
NeMo Gym / NeMo Evaluator; BF16; GPQA Diamond (no tools)
成绩出处Humanity's Last Exam厂商自报 · 越高越好
11.72%
NeMo Gym / NeMo Evaluator; BF16; text-only, no tools
成绩出处IFBench厂商自报 · 越高越好
71.88%
NeMo Gym / NeMo Evaluator; BF16; loose
成绩出处MMLU-Pro厂商自报 · 越高越好
81.94%
NeMo Gym / NeMo Evaluator; BF16
成绩出处OmniScience厂商自报 · 越高越好
17.5%
NeMo Gym / NeMo Evaluator; BF16; AA-Omniscience accuracy
成绩出处PinchBench厂商自报 · 越高越好
85.37%
NeMo Gym / NeMo Evaluator; BF16
成绩出处SciCode厂商自报 · 越高越好
32.6%
NeMo Gym / NeMo Evaluator; BF16
成绩出处SWE-bench Multilingual厂商自报 · 越高越好
39.33%
NeMo Gym / NeMo Evaluator; BF16
成绩出处SWE-Bench Verified厂商自报 · 越高越好
51.56%
NeMo Evaluator; BF16
成绩出处Tau3 Banking厂商自报 · 越高越好
9.28%
NeMo Gym / NeMo Evaluator; BF16; τ³-bench (Banking)
成绩出处Terminal-Bench 2.1厂商自报 · 越高越好
24.58%
NeMo Evaluator; BF16
成绩出处