chunklaya: why Laya stops reading after ~200 tokens, and a chunking harness that fixes it

Measures the decision head's attention drop-off on long inputs across 400 documents, then chunks, screens and aggregates, or locates the one passage that matters.

Open on GitHub

Preview of chunklaya: why Laya stops reading after ~200 tokens, and a chunking harness that fixes it

Launch: Laya
Use case: Data
Format: Repo
Source: github.com

Mukhamediyar

See all

mizorewww

@mizorewww · Sep 19

Validated Core ML ports of all Laya checkpoints. One short multilingual decision in 4.98 ms P50 on M3 Max with ANE FP16, at 2.78x better energy per decision than MLX FP16.

雨夹雪❄️

@mizorewww · Sep 20

The port announcement: Laya on MLX, under 1 GB RAM, choosing the snake’s next move 60 times a second on a local M3 Max.

atomic.chat

@atomic_chat_hq · Sep 21

Atomic.chat ran Laya on a 16 GB MacBook Air against Jev over the API on a Grok-built Tetris. Local won on speed: 11x faster decisions.

codejunkie99

@codejunkie99 · Sep 22

Open on GitHub

chunklaya: why Laya stops reading after ~200 tokens, and a chunking harness that fixes it

More built with Laya