Chen and Yin propose Auto-Diagnosis and Skill Discovery (ADSD), an automated framework that does not update model weights.
Diagnosis-first framework boosts solver performance across numerical domains
ADSD systematically identifies causes of poor solver behavior and uses that diagnosis to guide discovery of tailored numerical methods, packaged as reusable skills.
Chinese Tech
Peter Chen · Wotao Yin
University of California, Berkeley · Alibaba Group
Research Digest··3 min read
The authors introduce ADSD, a diagnosis-first framework that separates solver improvement into three stages: observing behavior, diagnosing underlying numerical conditions, and discovering targeted methods packaged as reusable skills.
Why this paper
From Alibaba Group and University of California, Berkeley
In one line
ADSD links numerical diagnosis to reusable solver skill discovery, improving solver accuracy, robustness, and efficiency across four numerical domains.
What we could check
- ·No code link found
- ·No weights link found
- ·No dataset link found
- ·No compute details found
- ✓Limitations stated by the authors
- ✓Reports numbers on named benchmarks
Observed from the paper text and links we have. Absence here means we did not find it, not that it does not exist.
§