Skip to content
View original post on X: BAAIOfficial
Published
24/100

LLaCO-8B reports top scores among diffusion-based LLMs on nine benchmarks

AIBAAI says its LLaCO-8B, a diffusion-based LLM, scores 62.12 on MATH, 47.02 points above DSL-LLaDA-8B's 15.10. The post reports 48.17 on HumanEval, 80.74 on GSM8K, 69.89 on MMLU and 31.25 on GPQA, and says these results beat discrete models including Dream-7B (51.20 on MATH) and LLaDA-8B (43.82).
BAAIVerified on XOfficial
@BAAIBeijing
View on X

Part of a thread · earlier post

Evaluated across 9 standard benchmarks, LLaCO-8B sets a new benchmark for diffusion-based LLMs:

📊 Crushes Continuous SOTA:
• +47.02 points on MATH over DSL-LLaDA-8B (62.12 vs. 15.10)
• +27.44 points on HumanEval (48.17 vs. 20.73)

🔥 Outperforms Discrete Giants:
• MATH: 62.12 (surpassing Dream-7B's 51.20 & LLaDA-8B's 43.82)
• GSM8K: 80.74 | MMLU: 69.89 | GPQA: 31.25

Source details

BAAI · x.comPublished

Original title: Evaluated across 9 standard benchmarks, LLaCO-8B sets a new benchmark for diffusion-based LLMs:

Topics

Data & trainingChina AIModel releasesEvals & benchmarks