Alibaba's SSR-GRPO: RL for E-commerce Dense Retrieval
_reachsumit · x · 2026-08-21
Alibaba presents SSR-GRPO, a reinforcement learning method for e-commerce search. To address bias from using LLMs as reward models, SSR-GRPO leverages Semantic IDs and dense vectors for unbiased relevance scoring. It also mines hard negatives to mask noisy samples, improving the handling of complex and implicit semantics.
More from Research
- Devin Oncall tops SREGym with 100% end-to-end success, auto-resolving production incidents — tianyin_xu · 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