cd /news/large-language-models/reading-between-dots-decoding-hidden… · home topics large-language-models article
[ARTICLE · art-64302] src=arxiv.org ↗ pub= topic=large-language-models verified=true sentiment=· neutral

Reading Between Dots: Decoding Hidden Computation Across Filler Tokens (ICML'26)

Researchers at ICML'26 found that frontier LLMs like DeepSeek V3 and Kimi K2 perform structured, legible hidden computation over filler tokens (e.g., dots) during multi-step reasoning, with attention routing and logit-lens readouts revealing intermediate values. An unsupervised decoding pipeline recovered these values with 80-95% accuracy, showing that hidden computation is readable from the residual stream even when surface tokens carry no reasoning trace.

read2 min views1 publishedJul 18, 2026
Reading Between Dots: Decoding Hidden Computation Across Filler Tokens (ICML'26)
Image: source
[Submitted on 3 Jul 2026]


[View PDF](/pdf/2607.03502)

[HTML (experimental)](https://arxiv.org/html/2607.03502v1)

Abstract:Frontier LLMs can perform multi-step reasoning over content-free filler tokens like dots or counting sequences, producing correct answers with no visible chain-of-thought (CoT). This is a limit case for behavioral oversight, where surface tokens carry no information about the underlying reasoning. But hidden from the output is not the same as hidden from us. On four task families (fact retrieval, parallel numeric composition, string manipulation, and in-context computation), two open-weights frontier models (DeepSeek V3, Kimi K2) compute over filler tokens in a structured, legible way: attention routes the question through the filler region to the answer, logit-lens readouts show retrieved facts emerging early and their composition crystallizing in late layers, and KV-cache transplants at filler positions causally swap outputs between examples. We introduce an unsupervised decoding pipeline that takes only hidden states as input and recovers intermediate values with 80-95% accuracy (best LLM judge) across both models and all four tasks, without ground-truth labels or training. Hidden computation that defeats behavioral CoT monitoring is, on these tasks, directly readable from the residual stream, suggesting monitorability is a property of the model's full computational trace, not just its surface tokens.

Current browse context:

cs.CL

References & Citations

...

Bibliographic Explorer

(What is the Explorer?) Connected Papers

(What is Connected Papers?) Litmaps

(What is Litmaps?) scite Smart Citations

(What are Smart Citations?)# Code, Data and Media Associated with this Article alphaXiv

(What is alphaXiv?) CatalyzeX Code Finder for Papers

(What is CatalyzeX?) DagsHub

(What is DagsHub?) Gotit.pub

(What is GotitPub?) Hugging Face

(What is Huggingface?) ScienceCast

(What is ScienceCast?)# Demos Influence Flower

(What are Influence Flowers?) CORE Recommender

(What is CORE?)# arXivLabs: experimental projects with community collaborators arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.

Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.

Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.

── more in #large-language-models 4 stories · sorted by recency
── more on @deepseek v3 3 stories trending now
sponsored brought to you by zahid.host 4,200+ EU-deployed projects
reading about agents? ship yours in a single git push.

Run your AI side-project on zahid.host

EU-based hosting, git-push deploys, automatic HTTPS, no cold starts. Free tier with a custom domain — perfect for shipping the agent you just read about.

$git push zahid main
Live at https://your-agent.zahid.host
Get free account → Pricing
from €0/mo · no card required
LIVE [news/reading-between-dots…] indexed:0 read:2min 2026-07-18 ·