Predicting Future Behaviors in Reasoning Models Enables Better Steering

• URL: http://arxiv.org/abs/2606.11172v1
• Date: Tue, 09 Jun 2026 17:49:24 GMT
• Title: Predicting Future Behaviors in Reasoning Models Enables Better Steering
• Abstract summary: Test-time steering controls large reasoning models (LRMs) by intervening on their hidden representations, but it can degrade output quality.<n>We argue that prior steering work implicitly relies on internal features that detect behavior in already generated text.<n>We train activation probes to predict future behavior likelihoods from intermediate reasoning steps.<n>These probes predict the most likely behavior with 64%-91% accuracy, revealing a separate type of internal prediction features.
• Score: 35.89008737242842
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Deployed large reasoning models (LRMs) often behave unexpectedly. Test-time steering controls LRM outputs by intervening on their hidden representations, but it can degrade output quality. We argue that prior steering work implicitly relies on internal features that detect behavior in already generated text. We show that these detection features are poor predictors of future behavioral outcomes, and thus not the natural intervention target. Instead, we train activation probes to predict future behavior likelihoods from intermediate reasoning steps. These probes predict the most likely behavior with 64%-91% accuracy, revealing a separate type of internal prediction features. Building on these prediction features, we introduce a text-level steering method, Future Probe Controlled Generation. FPCG samples multiple candidate sentences and chooses the best one according to a probe predicting the future behavior likelihood. This enables steering with almost no output quality degradation. FPCG also enables steering in several evaluations where activation steering fails. These results show that distinguishing detection and prediction features enables a more nuanced approach to controlling LRM behaviors.

Related papers

• PredVLA: Predictive Sensorimotor Modeling for Sub-Million-Parameter Robot Manipulation [4.55011976634859]
  We present PredVLA, a language-conditioned predictive-coding policy with only 0.68 million trainable network parameters and no robot-data pretraining.<n>Its hierarchical recurrent dynamics predict visual features and proprioception, while observations influence latent state only through prediction-error-driven online inference.
  arXiv  Detail & Related papers  (2026-08-27T06:27:11Z)
• When is Your LLM Steerable? [56.656180566692946]
  Activation steering offers a lightweight approach to control language models' behavior at inference time.<n>Finding the regime and boundaries of successful steering typically requires expensive grid searches.<n>We investigate whether steerability can be predicted from the model's internal states at the beginning of the generation process.
  arXiv  Detail & Related papers  (2026-06-10T02:55:34Z)
• Monitoring the Internal Monologue: Probe Trajectories Reveal Reasoning Dynamics [42.27928681606673]
  Large Reasoning Models (LRMs) introduce new opportunities for safety monitoring through their Chain of Thought (CoT) reasoning.<n>We investigate the hidden representations of LRMs to determine whether future behavior can be predicted from prompt and CoT representations.<n>We find that future model behavior is more distinguishable when examined over the full trajectory than from a single static prediction.
  arXiv  Detail & Related papers  (2026-05-18T15:29:04Z)
• Mechanisms of Introspective Awareness [4.6752132340946835]
  Recent work shows that LLMs can sometimes detect when steering vectors are injected into their residual stream and identify the injected concept.<n>But what mechanisms underlie this capability, and do they reflect genuine introspective circuitry or more shallow awarenesss?<n>We investigate these questions in open-source models and establish three main findings.
  arXiv  Detail & Related papers  (2026-03-22T20:45:34Z)
• From Data to Behavior: Predicting Unintended Model Behaviors Before Training [78.37660873165284]
  We introduce Data2Behavior, a new task for predicting unintended model behaviors prior to training.<n>We also propose Manipulating Data Features (MDF), a lightweight approach that summarizes candidate data through their mean representations.<n>Experiments on Qwen3-14B, Qwen2.5-32B-Instruct, and Gemma-3-12b-it confirm that MDF can anticipate unintended behaviors and provide insight into pre-training vulnerabilities.
  arXiv  Detail & Related papers  (2026-02-04T16:37:17Z)
• Language Models Can Predict Their Own Behavior [29.566208688211876]
  Language models (LMs) can exhibit specific behaviors,' such as a failure to follow alignment training, that we hope to detect and react to during deployment.<n>We show that probes trained on the internal representation of input tokens alone can predict a wide range of eventual behaviors over the entire output sequence.<n>An early warning system built on the probes reduces jailbreaking by 91%.
  arXiv  Detail & Related papers  (2025-02-18T23:13:16Z)
• Curb Your Attention: Causal Attention Gating for Robust Trajectory Prediction in Autonomous Driving [6.897286554827871]
  Trajectory prediction models in autonomous driving are vulnerable to perturbations from non-causal agents.<n>We propose $textitCausal tRajecTory predICtion$ $textbf(CRiTIC)$ to identify inter-agent causal relations over a window of past time steps.
  arXiv  Detail & Related papers  (2024-09-23T20:01:20Z)
• Model Predictive Control with Gaussian-Process-Supported Dynamical Constraints for Autonomous Vehicles [82.65261980827594]
  We propose a model predictive control approach for autonomous vehicles that exploits learned Gaussian processes for predicting human driving behavior. A multi-mode predictive control approach considers the possible intentions of the human drivers.
  arXiv  Detail & Related papers  (2023-03-08T17:14:57Z)
• Behavioral Intention Prediction in Driving Scenes: A Survey [70.53285924851767]
  Behavioral Intention Prediction (BIP) simulates a human consideration process and fulfills the early prediction of specific behaviors. This work provides a comprehensive review of BIP from the available datasets, key factors and challenges, pedestrian-centric and vehicle-centric BIP approaches, and BIP-aware applications.
  arXiv  Detail & Related papers  (2022-11-01T11:07:37Z)
• What Should I Know? Using Meta-gradient Descent for Predictive Feature Discovery in a Single Stream of Experience [63.75363908696257]
  computational reinforcement learning seeks to construct an agent's perception of the world through predictions of future sensations. An open challenge in this line of work is determining from the infinitely many predictions that the agent could possibly make which predictions might best support decision-making. We introduce a meta-gradient descent process by which an agent learns what predictions to make, 2) the estimates for its chosen predictions, and 3) how to use those estimates to generate policies that maximize future reward.
  arXiv  Detail & Related papers  (2022-06-13T21:31:06Z)
• Control-Aware Prediction Objectives for Autonomous Driving [78.19515972466063]
  We present control-aware prediction objectives (CAPOs) to evaluate the downstream effect of predictions on control without requiring the planner be differentiable. We propose two types of importance weights that weight the predictive likelihood: one using an attention model between agents, and another based on control variation when exchanging predicted trajectories for ground truth trajectories.
  arXiv  Detail & Related papers  (2022-04-28T07:37:21Z)
• On complementing end-to-end human motion predictors with planning [31.025766804649464]
  High capacity end-to-end approaches for human motion prediction have the ability to represent subtle nuances in human behavior, but struggle with robustness to out of distribution inputs and tail events. Planning-based prediction, on the other hand, can reliably output decent-but-not-great predictions.
  arXiv  Detail & Related papers  (2021-03-09T19:02:45Z)

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.