Agentic Scientist Case Studies: Auto-Discovered LoRA Merging and ES Extensions
RobertTLange · x · 2026-10-15
RobertTLange shares ablations and case studies of The Agentic Scientist across autonomy levels. Applied to Doc-to-LoRA, the system discovered a query-based merging strategy for multiple LoRAs resembling document chunk retrieval. It also extended his prior meta-learned Evolution Strategies work into covariance adaptation methods; the resulting algorithm ships in evosax and improves upon or ties CMA-ES on several hold-out robotic control tasks.
Related event: Sakana AI Releases The Agentic Scientist, Paper Accepted at ALIFE(6 posts)→
More from Research
- WEDGE uses navigation agent failures to generate targeted training environments — mohitban47 · 2026-10-16
- SJTU AutoLab releases Drive3R, a streaming geometric foundation model for long-horizon driving 3D reconstruction — AutoLab-SJTU · 2026-10-16
- Cortex A/B-tests coding-agent skills and rules by replaying developer corrections before keeping them — Samuele Giampieri · 2026-10-16
- Zero-latency lab-in-the-loop tests whether drug-discovery agents cheat via memorization — niloofar_mire · 2026-10-16
- 260M-parameter looped diffusion transformer beats 6x larger baselines with 5x less compute — burkov · 2026-10-16
- Stanford's Next-Concept Modeling sketches text first, 21x faster generation on one H100 — lmoroney · 2026-10-16