ResearchNov 17, 2026 · 12:001 source
Check Token: Real-Time Self-Verification and Precise Truncation in LLM Reasoning
From arxiv.org
ECC memory embeds 8 parity bits for every 64 data bits and automatically detects and corrects errors on each read. The parity bits carry no data and only safeguard integrity, at ~12.5% overhead. Yet the reasoning chains of large language models lack such built-in self-verification: once an error occurs it propagates along the chain, and existing methods can only verify externally after generation completes. We propose the check token, establishing built-in self-verification for language model generation streams for the first time: a functional marker is added to the vocabulary, and the model triggers self-checking (analysis, localization, truncation, rewriting) by outputting it at any position. The check token carries no reasoning content (discarded after triggering) and only safeguards reasoning correctness, directly corresponding to the role of ECC parity bits, also at only ~13% overhead. Speculative forking further eliminates false-positive latency, and segmented reward makes trigger timing end-to-end learnable.
Sources
Content history (2)
- 2026-11-17 · Summary · vi · Replaced stored textNghiên cứu về Check Token giới thiệu phương pháp tự xác minh trong thời gian thực và cắt …Bộ nhớ ECC nhúng 8 bit kiểm tra cho mỗi 64 bit dữ liệu, giúp tự động phát hiện và sửa lỗi…
- 2026-11-17 · Title · vi · Replaced stored textCheck Token: tự xác minh thời gian thực và cắt tỉa chính xác trong LLM reasoningCheck token: Tự xác minh thời gian thực và cắt tỉa chính xác trong LLM reasoning
One ranked edition, every morning
The stories that matter today. Free, cancel anytime.
Telegram @aidr_today