Sentence Transformers v6.3 ships token pooling-aware training for leaner multi-vector retrieval
antoine_chaffin · x · 2026-11-12
tomaarsen released Sentence Transformers v6.3 with a key new feature: token pooling-aware training. It lets you train multi-vector (late interaction) models to preserve retrieval quality while storing far fewer document vectors. Pooling-aware training of late interaction models is now officially supported, with an accompanying blog post and thread explaining the approach.
More from Research
- ArcGrav: a cross-model Agent Control Language to cut token overhead — ArchitecturalGravity · 2026-11-12
- Open-source dual-arm robot solves tangram puzzles with 90 demos, ~60% success — DanGrover · 2026-11-12
- FutureHouse's deslopper trains Kosmos to write indistinguishably from human experts — proteinrosh · 2026-11-12
- Fleuret amplifies debate: AI can't slash wet-lab experiments in unmapped biology — francoisfleuret · 2026-11-12
- Recursive's Eureka Machine ships Flint: a sparse 40B model with 2.5B active for local agentic coding — agihouse_org · 2026-11-12
- Recursive bets on open-endedness: agents running massive parallel experiments like scientist communities — agihouse_org · 2026-11-12