It Predicts Its Answer Length First [LLM Hidden States]

It knows its answer's length before word one A frozen language model, read by the dumbest tool in interpretability, turns out to know roughly how long its whole answer will be — before it writes a single word. But when the paper's most jaw-dropping scene turns out to be shot in the exact spot where its instruments are most broken, the real question becomes whether the model actually uses that number or just carries it. A clean fight over what counts as a 'plan' inside a next-word predictor. Full episode page: https://paperdive.ai/episodes/204-how... Paper: How Much is Left? LLMs Linearly Encode Their Remaining Output Length Read the paper: https://arxiv.org/abs/2607.05316 What you'll take away: Why a linear probe — a weighted sum too simple to compute — proves information was already written into the model's state rather than derived by the reader The three-predictor design (lazy forecaster, seeded countdown, full probe) that isolates exactly when length information appears and whether it gets revised mid-answer How a one-directional transfer matrix rules out 'the probe just memorized dataset quirks' and points to a general length direction The retraction spike: the probe's estimate leaping from ~71 to ~277 at the moment a model writes 'Wait — that can't be right' Why that showcase scene is weakest evidence — pulled from the probe's failure pile, only 5 examples, no control, absolute numbers 'frankly garbage' The core unproven claim: presence of the number is established three ways, but nobody has shown the model actually reads it when deciding to stop Chapters: 00:00 llm knows answer length 01:40 statistical drift null hypothesis 02:23 linear probes hidden states 04:27 lazy forecaster vs full probe 05:45 length encoded before generation 07:22 probe transfer matrix generalization 09:04 retraction token length jump 11:44 presence versus causal use 13:15 does model use length plan This episode is AI-generated. The script was written by an AI language model and the host voices were synthesized by Eleven Labs. The producer is not affiliated with Anthropic or Eleven Labs. The on-screen illustrations were generated by OpenAI GPT Image and Google Gemini. #AI #MachineLearning #LinearProbing