AI
ZooWork-ShopRanker: An Open, Preference-Aligned E-Commerce Reranker
Publication: Hugging Face Daily Papers
Author: Not available in collected metadata
Source published: 24/09/2026, 20:00:00 UTC
Briefing created: 28/09/2026, 06:35:10 UTC
Based on a feed excerpt; full article content has not been acquired. Source evidence retrieved 28/09/2026, 06:18:24 UTC.
Briefings
ZooWork-ShopRanker is a family of open e-commerce rerankers trained on preference-aligned data. The 8B model serves as a distillation teacher for the 4B and 0.6B models, which are then sharpened on judged pairs. The system uses a panel of reasoning LLMs to generate judge-labeled shopping preference pairs, with position-debiased judgments and agreement tiers. The 8B flagship model significantly outperforms open rerankers, beats its own un-aligned base, and the 0.6B model beats its size peer. Results hold in both text formats and extend to common MTEB benchmarks.
Key points
- The ZooWork-ShopRanker family consists of three models: 0.6B, 4B, and 8B, all trained on preference-aligned data using a panel of reasoning LLMs as a preference oracle.
- The 8B flagship model significantly outperforms open rerankers and beats its own un-aligned base, while the 0.6B model beats its size peer in both text formats and MTEB benchmarks.
Context and limitations
The benchmark relies on ~10,000 private-traffic preference pairs, which may not represent general web retrieval transfer performance. Results are limited to the specific formats and models tested in the study.
Generated by Bittrees-hosted qwen35-2b sha256:aaf42c8b7c3c on 28/09/2026, 06:35:10 UTC. Based on publisher-provided feed evidence; may not cover the full original article. Check the source for details.
View archived briefing on IPFS
bafybeigtwtdorzenie3ewsp5bd47jbryupezkbhhwfd37pbxehgoy5mi5q