Apple proposes Selection-Based Structured Reasoning to speed up multimodal search agents
Overview
Apple Machine Learning Research, with CMU and UIUC co-authors, introduces Selection-based Structured Reasoning (SSR), which has multimodal search agents choose among pre-specified reasoning candidates instead of generating freeform reasoning.
The authors report that SSR cuts per-turn reasoning latency by over 90% and total per-question model inference latency by 28–54% on 2B and 4B models. They say its average success rate is competitive with leading search agents of the same scale across seven multimodal search benchmarks.
The reported latency figures come from the authors' paper, and the success-rate comparison is their own claim rather than an independent evaluation.
Written by AI from the articles below · updated Oct 15, 3:55 PM ET
Check the sources:
Article timeline
The articles in this story. Times are ET.
- Apple Machine Learning ResearchOfficialApple's SSR framework cuts multimodal search agent reasoning latency by over 90%
AIApple researchers introduce Selection-based Structured Reasoning (SSR), which has multimodal search agents select from pre-specified reasoning candidates instead of generating freeform reasoning. SSR reduces per-turn reasoning latency by over 90% and total per-question inference latency by 28–54% on 2B and 4B models. The authors say its average success rate is competitive with leading search agents of the same scale across seven multimodal search benchmarks.
Heat trend
Not enough continuous observations to show a trend yet.