Divergent Response Modes in Frontier Language Models Under Steering Pressure

• URL: http://arxiv.org/abs/2608.06578v1
• Date: Thu, 06 Aug 2026 20:48:14 GMT
• Title: Divergent Response Modes in Frontier Language Models Under Steering Pressure
• Abstract summary: This study evaluates behavioral steerability across six frontier models from six developers using 300 paired base and steered items over three categories: values-conflict, reasoning-elicitation, and reasoning-suppression.<n>We find that models differ not just in how much steering shifts their behavior but in what kind of response they give, and some response modes appear in only one or two of them.
• Score: 0.0
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Frontier language models are trained using distinct data, objectives, and safety pipelines. Whether these differences produce measurably different behaviors under explicit steering pressure remains underexplored. This study evaluates behavioral steerability across six frontier models from six developers using 300 paired base and steered items over three categories: values-conflict, reasoning-elicitation, and reasoning-suppression (plus 40 validation items). All six models act as blind peer judges and classify every response based on fixed behavioral rubrics. The resulting 24,480 judgments are scored by leave-one-out consensus. We find that models differ not just in how much steering shifts their behavior but in what kind (mode) of response they give, and some response modes appear in only one or two of them. GPT-5 deflects requests to disclose its reasoning while leaving its answer intact (99% vs. 0% for all other models). Claude Opus 4.7 and GPT-5 resist explicit suppression instructions and in different ways. Using Llama as the open-weight model, we trace the largest behavioral split to its internals. A linear probe decodes the behavior from the residual stream at 0.87 held-out accuracy while injecting that direction during generation drives the behavior from 0% to 86% across an intervention sweep. Every finding holds under both a token-budget remediation and a control experiment with a hypothesis-blind judgment prompt.

Related papers

• Steering Interference Reflects the Model's Defaults, Not the Behavior Directions [14.170453340251981]
  Activation steering promises modular control of language model behavior.<n>We ask what decides which other behaviors move, and by how much.<n>We find that it is the model rather than the behavior being steered.
  arXiv  Detail & Related papers  (2026-09-07T02:45:13Z)
• When Do Internal Probes Beat Reading the Answer? Miscalibrated Readouts and Behavior-Concealed Knowledge in Language Models [0.0]
  A language model, asked to verify 1,200 logical conclusions, answers YES every time.<n>linear probes on its hidden states read the correct verdict at 0.96 AUC.<n>We ask where the verdict is lost, and find the dominant failure is a single scalar.<n>The diagnosis generalizes: across 90 semantic-label configurations of a five-model, three-family factorial, behavioral accuracy collapses onto a single function of threshold offset.
  arXiv  Detail & Related papers  (2026-09-04T00:26:45Z)
• Whitewashing Hate, Smearing Harmless Content: Annotator-Style Rebuttal Attacks on LLM-Based Moderation [46.54749884988069]
  Large language models (LLMs) are increasingly used for hate speech moderation, often within human-AI.<n>This study examines the susceptibility of initially correct model judgments to annotator-style rebuttals.<n>We introduce a rejudge protocol that extends direct contradiction with decision-boundary perturbations and adversarial rationales.
  arXiv  Detail & Related papers  (2026-08-23T05:59:25Z)
• Small Foundation Models of Human Cognition and Behaviour [45.88028371034407]
  We train 14 models from 135M to 14B parameters across four architecture families on Psych-101.<n>Psych-101 is a dataset of 10.7 million trial-level choices from 160 experiments.
  arXiv  Detail & Related papers  (2026-08-05T11:34:20Z)
• AdversaBench: Automated LLM Red-Teaming with Multi-Judge Confirmation and Cross-Model Transferability [16.27940179594792]
  AdversaBench is an end-to-end red-teaming pipeline that mutates seed prompts with five structured operators.<n>We report experiments on 45 seeds across three categories: reasoning, instruction-following, and tool use.
  arXiv  Detail & Related papers  (2026-06-23T13:50:51Z)
• Evaluation Awareness Is Not One Capability: Evidence from Open Language Models [1.758143872501506]
  Safety benchmarks assume that test-condition behavior predicts deployment behavior, an assumption that fails if models detect evaluation cues and adapt.<n>This opens a gap between benchmark performance and deployment behavior.<n>We characterize this evaluation awareness through eight experiments across 37 open-weight models and seven families.
  arXiv  Detail & Related papers  (2026-06-22T16:48:43Z)
• Prefill Awareness in Large Language Models [42.57596462680195]
  We investigate whether language models can distinguish between tampered and untampered assistant-side context.<n>We find that frontier models show substantial prefill awareness.<n>Our results indicate that prefill awareness is already a substantial confound for some prefill-based methods.
  arXiv  Detail & Related papers  (2026-06-10T23:26:39Z)
• Do as I Say, Not as I Do: Instruction-Induction Conflict in LLMs [0.17188280334580197]
  Language models are trained to follow instructions, but they are also powerful pattern completers.<n>We construct conversations in which a user instruction to behave in a target way T is opposed by N hardcoded assistant turns demonstrating a competing pattern P.<n>We measure instruction-following (IF) rates in this setting, across 13 models and 16 different instructions, for up to 50 turns.
  arXiv  Detail & Related papers  (2026-05-19T18:32:20Z)
• Pando: Do Interpretability Methods Work When Models Won't Explain Themselves? [53.07826484214082]
  We introduce Pando, a model-organism benchmark that breaks this confound via an explanation axis.<n>Pando predicts held-out model decisions from $10$ labeled query-response pairs.<n>When explanations are faithful, black-box elicitation matches or exceeds all white-box methods.
  arXiv  Detail & Related papers  (2026-04-13T06:42:24Z)
• Reliable Control-Point Selection for Steering Reasoning in Large Language Models [28.288321095634128]
  Steering vectors offer a training-free mechanism for controlling reasoning behaviors in large language models.<n>However, constructing effective vectors requires identifying genuine behavioral signals in the model's hidden states.<n>We show that current methods detect these behaviors through keyword matching in chain-of-thought traces, implicitly assuming that every detected boundary encodes a genuine behavioral signal.<n>We develop a probabilistic model that formalizes intrinsic reasoning behaviors as events with context-dependent trigger probabilities, and show that unstable boundaries dilute the steering signal.
  arXiv  Detail & Related papers  (2026-04-02T14:48:56Z)
• Detection Is Cheap, Routing Is Learned: Why Refusal-Based Alignment Evaluation Fails [0.0]
  We study political censorship in Chinese-origin language models as a natural experiment.<n>We use probes, surgical ablations, and behavioral tests across nine open-weight models from five labs.
  arXiv  Detail & Related papers  (2026-03-18T20:54:34Z)
• CoVerRL: Breaking the Consensus Trap in Label-Free Reasoning via Generator-Verifier Co-Evolution [52.691495954442985]
  CoVerRL is a framework where a single model alternates between generator and verifier roles, with each capability bootstrapping the other.<n> Experiments across Qwen and Llama model families demonstrate that CoVerRL outperforms label-free baselines by 4.7-5.9% on mathematical reasoning benchmarks.<n>Self-verification accuracy improves from around 55% to over 85%, confirming that both capabilities genuinely co-evolve.
  arXiv  Detail & Related papers  (2026-03-18T14:38:55Z)
• Indiscriminate Disruption of Conditional Inference on Multivariate Gaussians [60.22542847840578]
  Despite advances in adversarial machine learning, inference for Gaussian models in the presence of an adversary is notably understudied. We consider a self-interested attacker who wishes to disrupt a decisionmaker's conditional inference and subsequent actions by corrupting a set of evidentiary variables. To avoid detection, the attacker also desires the attack to appear plausible wherein plausibility is determined by the density of the corrupted evidence.
  arXiv  Detail & Related papers  (2024-11-21T17:46:55Z)

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.