FreshStack Refreshes RAG Leaderboard, Adds Cohere Embed v5 and New Retrieval Models
beirmug · x · 2026-10-23
The FreshStack team revamped its website and leaderboard, adding recent models including Cohere Embed v5, Google's EmbeddingGemma2, and Perplexity's late-interaction models. FreshStack is a RAG benchmark framework from Waterloo and Databricks (Nandan Thakur, Jimmy Lin, Omar Khattab et al.), building realistic retrieval evaluations from community Q&A across five niche technical domains; accepted at NeurIPS 2025 Datasets & Benchmarks and honored at BCS Search Solutions 2025. It's now officially included in RTEB (the newer MTEB) and adopted in Databricks' KARLBench, with an EMNLP 2026 Findings paper on technical drift in LangChain docs. The leaderboard ranks retrievers by average Recall@50 across dense, lexical, hybrid, and late-interaction setups.
Related event: FreshStack Retrieval Leaderboard Updated: Oten-Embedding-8B Takes Top Spot(2 posts)→
More from Research
- New paper makes 9 falsifiable predictions toward a major AI-enabled terror attack by 2027-2028 — StephenLCasper · 2026-10-23
- HumanoidPose recovers humanoid robots' full-body joint states from RGB video alone — cholz · 2026-10-23
- Sophont's Holon-S1 predicts genes and biomarkers from routine pathology slides — iScienceLuvr · 2026-10-23
- Glasgow study: raised eyebrows and slow blinks make speakers look confident — Limor_Raviv · 2026-10-23
- IFM open-sources RL360, the agentic RL infra behind K2 Horizon's training — rupspace · 2026-10-23
- Wayfarer discovers RL options from a single experience stream, SOTA on hard Atari games — MarlosCMachado · 2026-10-23