How Much is Left? LLMs Linearly Encode Their Remaining Output Length

• URL: http://arxiv.org/abs/2607.05316v1
• Date: Mon, 06 Jul 2026 16:56:04 GMT
• Title: How Much is Left? LLMs Linearly Encode Their Remaining Output Length
• Abstract summary: Large language models generate one token at a time, yet their responses show remarkably consistent length structure.<n>We ask whether the model carries an internal estimate of how much response remains.<n>We train minimal-capacity linear probes on frozen hidden states of three open-weight 7-8B models across seven completion-style datasets.
• Score: 3.541590945326217
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: Large language models generate one token at a time, yet their responses show remarkably consistent length structure: step-by-step solutions converge in predictable token counts, retrievals stop after a few sentences, retractions extend responses by measurable amounts. We ask whether the model carries an internal estimate of how much response remains. Training minimal-capacity linear probes on frozen hidden states of three open-weight 7-8B models across seven completion-style datasets, we find three converging pieces of evidence. First, total response length is linearly decodable from the prompt's last hidden state alone, before any output is emitted. Second, probe directions trained on natural-language datasets transfer broadly, including to controlled synthetic completions never seen in training, outperforming a statistical baseline; the converse direction generally fails, and this asymmetry is itself informative. Third, on curated high-loss completions, the probe's per-position estimate shifts upward at the moment the model retracts and restarts a partial solution, a directional behavior no position-only predictor can reproduce (qualitative, not aggregate). We frame this as approximate estimation of remaining generation length, distinct from exact-counting impossibility results for transformers, and interpret it as evidence that LLMs maintain a plan-like internal representation of output length (decodable, not necessarily used causally).

Related papers

• Bypass Observation: A Conceptual Design of a Non-Intrusive Layer-Wise Semantic Extraction Architecture [2.8943395720203386]
  We introduce Bypass Observation, a non-intrusive layer-wise readout architecture.<n>We derive a closed-form overhead approximation governed primarily by V/(12d)<n>We argue that Bypass Observation can make model iteration more observable while remaining only a partial, potentially misleading projection of hidden states.
  arXiv  Detail & Related papers  (2026-09-12T08:35:45Z)
• Wrong Prediction, Right Answer: Recovering Evidence from Collapsed LLM Sequence Scores [31.709549159768727]
  We observe a consistent readout gap across diverse reasoning benchmarks.<n> hidden-state probes successfully decode correct answers even when native sequence scoring completely collapses.<n>Our results show that many apparent zero-shot reasoning deficits are expression failures masking intact internal logic.
  arXiv  Detail & Related papers  (2026-08-31T16:43:55Z)
• TriLens: Per-Layer Logit-Lens Entropy for White-Box Hallucination Detection [22.258009855879752]
  We introduce TriLens, a detector for detecting hallucinations in language models.<n>At every layer, it reads the multi-head self-attention output, the feed-forward output, and the residual stream through the model's own logit lens.<n>The resulting 3L-dimensional trajectory describes how certainty forms across depth and across modules.
  arXiv  Detail & Related papers  (2026-05-31T05:48:21Z)
• A2RBench: An Automatic Paradigm for Formally Verifiable Abstract Reasoning Benchmark Generation [59.98516959731531]
  Abstract reasoning ability reflects intelligence and capacity of LLMs to extract and apply abstract rules.<n>Existing benchmarks either rely on expensive manual annotation, limiting their scale, or risk measuring memorization rather than genuine reasoning.<n>We introduce an automated pipeline named A2RBench, encompassing generation, expansion, evaluation, and analysis.
  arXiv  Detail & Related papers  (2026-05-17T06:14:20Z)
• When Does a Language Model Commit? A Finite-Answer Theory of Pre-Verbalization Commitment [6.48455847613423]
  Language models often generate reasoning before giving a final answer, but the visible answer does not reveal when the model's answer preference became stable.<n>We study this question through a narrow computable object: emphfinite-answer preference stabilization.<n>For a model state and specified answer verbalizers, we project the model's own continuation probabilities onto a finite answer set.<n>In binary tasks this yields an exact log-odds code, $()=S_(mathrmyesmid)-S_(mathrmnomid
  arXiv  Detail & Related papers  (2026-05-07T08:34:59Z)
• Probing the Trajectories of Reasoning Traces in Large Language Models [4.599673637363014]
  We propose a protocol to probe the trajectories of reasoning traces in large language models.<n>We find that accuracy and decision commitment consistently increase as the percentage of provided reasoning tokens grows.<n>We show that trajectory probing provides diagnostics for efficient and safer deployment of reasoning models.
  arXiv  Detail & Related papers  (2026-01-30T16:45:16Z)
• Temporal Predictors of Outcome in Reasoning Language Models [0.0]
  Chain-of-thought (CoT) paradigm uses the elicitation of step-by-step rationales as a proxy for reasoning.<n>We show that, for harder questions, a drop in predictive accuracy highlights a selection artifact.<n>Overall, our results imply that for reasoning models, internal self-assessment of success tends to emerge after only a few tokens.
  arXiv  Detail & Related papers  (2025-11-03T08:57:18Z)
• Why Can't Transformers Learn Multiplication? Reverse-Engineering Reveals Long-Range Dependency Pitfalls [54.57326125204404]
  Language models are increasingly capable, yet still fail at a seemingly simple task of multi-digit multiplication.<n>We study why, by reverse-engineering a model that successfully learns multiplication via emphimplicit chain-of-thought'
  arXiv  Detail & Related papers  (2025-09-30T19:03:26Z)
• Diffusion Language Models Know the Answer Before Decoding [56.96815863705218]
  Diffusion language models (DLMs) have emerged as an alternative to autoregressive approaches.<n>Our work highlights and leverage an overlooked property of DLMs early answer convergence.<n>We introduce Prophet, a training-free fast decoding paradigm that enables early commit decoding.
  arXiv  Detail & Related papers  (2025-08-27T15:40:25Z)
• CASTILLO: Characterizing Response Length Distributions of Large Language Models [3.5041586868397854]
  We introduce CASTILLO, a dataset characterizing response length distributions across 13 widely-used large language models.<n>Our analysis reveals significant inter- and intra-model variability in response lengths, as well as model-specific behaviors and occurrences of partial text degeneration in only subsets of responses.
  arXiv  Detail & Related papers  (2025-05-22T16:35:33Z)
• On the Transferability of Pre-trained Language Models: A Study from Artificial Datasets [74.11825654535895]
  Pre-training language models (LMs) on large-scale unlabeled text data makes the model much easier to achieve exceptional downstream performance. We study what specific traits in the pre-training data, other than the semantics, make a pre-trained LM superior to their counterparts trained from scratch on downstream tasks.
  arXiv  Detail & Related papers  (2021-09-08T10:39:57Z)
• Reflective Decoding: Beyond Unidirectional Generation with Off-the-Shelf Language Models [63.808843089941405]
  Large pretrained LanguageModels (LMs) generate text with remarkable quality, but only sequentially from left to right. We present Reflective Decoding, a novel unsupervised algorithm that allows for direct application of unidirectional LMs to non-sequential tasks. Our 2-step approach requires no supervision or even parallel corpora, only two off-the-shelf pretrained LMs in opposite directions.
  arXiv  Detail & Related papers  (2020-10-16T18:02:07Z)
• Information-Theoretic Probing with Minimum Description Length [74.29846942213445]
  We propose an alternative to the standard probes, information-theoretic probing with minimum description length (MDL) With MDL probing, training a probe to predict labels is recast as teaching it to effectively transmit the data. We show that these methods agree in results and are more informative and stable than the standard probes.
  arXiv  Detail & Related papers  (2020-03-27T09:35:38Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.