PulseAugur
EN
LIVE 05:40:24

Adaption Labs' AutoScientist tunes models 35% better than researchers

Adaption Labs has developed AutoScientist, an automated system designed to optimize AI model training. This system reportedly outperforms the original lab researchers in tuning models, achieving a 35% improvement on internal evaluations. AutoScientist achieves this by co-optimizing both the training data and the training recipe, while also incorporating self-improvement capabilities from previous runs. AI

IMPACT This automated tuning system could significantly accelerate AI development cycles and improve model performance across various applications.

RANK_REASON Product launch of a novel AI system with claimed performance improvements. [lever_c_demoted from significant: ic=1 ai=1.0]

Read on Mastodon — fosstodon.org →

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →

COVERAGE [1]

  1. Mastodon — fosstodon.org TIER_1 English(EN) · [email protected] ·

    Can an automated system tune models better than the lab that built it? Adaption Labs says yes: its AutoScientist co-optimizes data and training recipe, self-imp

    Can an automated system tune models better than the lab that built it? Adaption Labs says yes: its AutoScientist co-optimizes data and training recipe, self-improves over past runs, and beats the lab's researchers by 35% on its own in-house evals. https:// benjaminhan.net/posts/2…