Sentence Transformers adds tunable compression: R2 rerankers keep 92% nDCG@10 at 128x
tomaarsen · x · 2026-10-30
Sentence Transformers maintainer tomaarsen shares usage tips for the KaLM R2 reranker models:
- Set model[0].chunksize = 128 in Sentence Transformers for stronger compression, or None to disable pooling.
- The authors report all three R2 models retain over 92% of average nDCG@10 at 2x compression on both BEIR and MIRACL.
- The compression ratio is adjustable to trade reranking quality for compute.
- Caveat: the CrossEncoder path re-encodes documents on each call, so the opening example doesn't demonstrate offline document-caching savings.
Related event: KaLM R2 Reranker Open-Sourced, Retains 92% Recall at 128x Compression(2 posts)→
More from coding & agent
- Simlock: open-source control plane gives parallel coding agents their own iOS/Android simulators — Scobleizer · 2026-10-30
- Keyfleet launches open canvas where personal AI agents team up with a shared USDC treasury — seanwbren · 2026-10-30
- Clear Thought MCP Server adds structured reasoning and mental models to AI clients — modelcontextprotocol · 2026-10-30
- NSE Stock Insights MCP brings AI analysis to India's Nifty 500 stocks — modelcontextprotocol · 2026-10-30
- OpenRouter rebuilds provider routing: 46% fewer capacity errors, 16.9% faster TTFT — steipete · 2026-10-30
- Databricks launches Agent Memory, tripling agent pass rates with 31% fewer tokens — matei_zaharia · 2026-10-30