One int8 op makes exact rescoring nearly free in PLAID-style late-interaction search
antoine_chaffin · x · 2026-09-03
next-plaid 1.7.0 turns the slow half of PLAID-style late-interaction search — exact rescoring — into nearly free using a single int8 operation. The benchmark comparison shows the "decompress" segment is gone versus 1.6.5, it works on existing indexes, and runs CPU-only. Details in the author's thread.
More from Infra
- Open-source ArtSmoker pipeline turns text prompts into textured, Blender-ready 3D models on your own AWS — niravdd · 2026-09-03
- KV cache, not parameter count, may be the real bottleneck for long-context local models — jonejy · 2026-09-03
- RX 6800 XT vs RX 9060 XT for ComfyUI: is the $105 premium worth it for LTX and Wan? — Nice-Regret-9207 · 2026-09-03
- Developers warn a new Cloudflare feature could hurt your SEO — turn it off — gaganghotra_ · 2026-09-03
- If Amazon Trainium is any good, why weren't they at Hot Chips? — firstadopter · 2026-09-03
- Dragonfly rethinks Redis with sharded multi-threading to scale across modern multi-core servers — techNmak · 2026-09-03