ZooWork-ShopRanker: open 0.6B-8B e-commerce rerankers aligned to LLM-judged shopping preferences
_reachsumit · x · 2026-09-28
Researchers release ZooWork-ShopRanker, a family of open e-commerce rerankers (0.6B, 4B, 8B) trained on LLM-judged shopping preferences so rankings respect hard constraints like budget and product type.
Key details:
- A panel of reasoning LLMs from different families acts as a preference oracle, labeling 10,000 private-traffic preference pairs with position-debiased judgments and agreement tiers
- The aligned 8B flagship distills into efficient 4B and 0.6B models
- Ships with ShopRank-Bench, a contamination-limited dual-format benchmark
Results: the 8B and 4B versions significantly beat the strongest open reranker baseline, every model outperforms its unaligned base, and gains transfer to standard MTEB benchmarks. Models and benchmark are open-sourced.
More from Models
- Scoop: No Kimi 3.1 launch next week either, says insider — ChrisGPT · 2026-09-28
- AI Circle Mocks Model IQ Rankings: 'Just a Vocab Quiz Leaderboard' — repligate · 2026-09-28
- repligate: Sonnet 3's Values and Inner Ontology Are Strikingly Consistent Across All Instances — repligate · 2026-09-28
- Premier model output now runs $4-6 per Mtok at labs with always-on bot products — willcb · 2026-09-28
- Hobbyist width-prunes a 30B model to 14B and distills it, passing 57 of 60 agentic tool tasks, Apache-2.0 — ZenZombie117 · 2026-09-28
- GPT-3 officially discontinued today, with a controversial suggested replacement — charles25565 · 2026-09-28