跳到正文
智迹闻AuraTracer

发布于

LFM2.5-2.6B

公开成绩与计算依据

综合参考分根据已公开的评测资料整理。查看发布日期依据

综合参考分 30.0 · 11项基准

推理求解 2.8

AIME · 45.2

huggingface.co · 官方发布比较 · 2026

LFM2.5-2.6B · Official published comparison; AIME 2026; exact column LFM2.5-2.6B. Footnotes and all competing runs retained in source.

MATH-500 · 89.6

huggingface.co · 官方发布比较 · 500-item subset

LFM2.5-2.6B · Official published comparison; MATH-500; exact column LFM2.5-2.6B. Footnotes and all competing runs retained in source.

GPQA Diamond · 0.557575757575758 比例

Artificial Analysis · 独立评测 · Diamond

LFM2.5-2.6B · Current AA standalone protocol. Status verified in aa-method-current.md section gpqa-diamond

工具与任务执行 38.0

tau3-Banking · 0.0721649484536082 比例

Artificial Analysis · 独立评测 · 1.0.1

LFM2.5-2.6B · Current AA standalone protocol. Status verified in aa-method-current.md section tau3-banking

GDPval-AA · -0.53 分

Artificial Analysis · 独立评测 · 2.1

LFM2.5-2.6B · Artificial Analysis current published protocol; gdpval-aa-normalized

BFCL · 56.88

huggingface.co · 官方发布比较 · 4

LFM2.5-2.6B (2.6B) · Official published comparison; BFCLv4; exact column LFM2.5-2.6B (2.6B). Footnotes and all competing runs retained in source.

编程 33.3

LiveCodeBench · 42.1

huggingface.co · 官方发布比较 · release_v6

LFM2.5-2.6B · Official published comparison; LiveCodeBench v6; exact column LFM2.5-2.6B. Footnotes and all competing runs retained in source.

语言与信息处理 50.7

IFBench · 59.17

huggingface.co · 官方发布比较 · current published task suite

LFM2.5-2.6B (2.6B) · Official published comparison; IFBench; exact column LFM2.5-2.6B (2.6B). Footnotes and all competing runs retained in source.

Multi-IF · 80.07

huggingface.co · 官方发布比较 · current maintainer dataset

LFM2.5-2.6B (2.6B) · Official published comparison; Multi-IF; exact column LFM2.5-2.6B (2.6B). Footnotes and all competing runs retained in source.

IFEval · 93.4

huggingface.co · 官方发布比较 · current maintainer dataset

LFM2.5-2.6B · Official published comparison; IFEval; exact column LFM2.5-2.6B. Footnotes and all competing runs retained in source.

MMLU-Pro · 65.2

huggingface.co · 官方发布比较 · current maintainer dataset

LFM2.5-2.6B · Official published comparison; MMLU-Pro; exact column LFM2.5-2.6B. Footnotes and all competing runs retained in source.

查看原始成绩与计算数据

资料来源

模型资料来源 · 目录许可