A Low-Rank Subspace Analysis of LLM Interventions

• URL: http://arxiv.org/abs/2606.14388v1
• Date: Fri, 12 Jun 2026 12:24:25 GMT
• Title: A Low-Rank Subspace Analysis of LLM Interventions
• Abstract summary: Interventions designed to modify a particular behavior in LLMs often produce unintended changes in other behaviors.<n>This lack of targeted control makes it difficult to design and implement reliable safety controls.<n>We model behaviors as low-rank subspaces in activation space, and study how interventions influence across behaviors.
• Score: 21.607167713433622
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: Interventions designed to modify a particular behavior in LLMs, such as refusal or sycophancy, often produce unintended changes in other behaviors. This lack of targeted control makes it difficult to design and implement reliable safety controls. To understand these side-effects, we introduce a diagnostic framework for analyzing interacting behaviors in LLMs. We model behaviors as low-rank subspaces in activation space, and study how interventions influence across behaviors. Across multiple instruction-tuned models (7B-70B) and across refusal, jailbreak, and sycophancy settings, we find that different behaviors share internal representations, and intervening on one behavior alters others in asymmetric ways. Some behaviors act as upstream control points whose interventions propagate broadly across other behaviors, while others remain more isolated. We relate these effects to two geometric quantities: (i) the overlap between behavior subspaces, measured as the average squared cosine of principal angles, and (ii) the angle between each behavior subspace and the decision subspace (capturing the model's final decision e.g., refuse vs. comply). Empirically, intervention effects on other behaviors tend to be larger for behavior pairs with higher subspace overlap, and for source behaviors whose subspaces lie closer (smaller angle) to the decision subspace. These findings highlight a challenge for targeted behavior control: behaviors are difficult to modify independently, as interventions can propagate through shared representations and asymmetric interactions.

Related papers

• Locating and Controlling Implicit Personalization in Large Language Models [13.786400660073065]
  We show that a localized internal activation signal tracks changes in recommendations, with correlations up to r=0.87.<n>We show that removing the internal signal associated with one cue can suppress its influence, often more effectively than asking the model to ignore demographics via prompting.<n>These results connect implicit personalization behavior to an internal signal that can be analyzed and causally controlled.
  arXiv  Detail & Related papers  (2026-08-12T07:18:31Z)
• Your LLM, Your Style: Behavioral Mode Axes for LLM Behavioral Control [15.236237731175192]
  Large language models (LLMs) increasingly act in interactive settings where their behavioral styles affect user experience, safety, and downstream decision making.<n>Existing LLM personality studies largely rely on self-report questionnaires administered in first-person settings.<n>We introduce a situated behavioral-data framework for studying and controlling LLM behavioral personality.
  arXiv  Detail & Related papers  (2026-08-11T09:25:47Z)
• The Illusion of Intervention: Your LLM-Simulated Experiment is an Observational Study [56.649987197786096]
  Large language models (LLMs) show potential as simulators of human behavior.<n>LLMs can induce unintended shifts in latent user attributes, causing user drift.<n>We formalize the confounding or selection bias that can arise due to user drift.
  arXiv  Detail & Related papers  (2026-05-20T06:09:41Z)
• On the Structural Non-Preservation of Epistemic Behaviour under Policy Transformation [51.56484100374058]
  We formalise such information-conditioned interaction patterns as behavioural dependency.<n>This induces a probe-relative notion of $$-behavioural equivalence and a within-policy behavioural distance.<n>Results identify structural conditions under which probe-conditioned behavioural separation is not preserved under common policy transformations.
  arXiv  Detail & Related papers  (2026-02-24T22:55:21Z)
• Why Steering Works: Toward a Unified View of Language Model Parameter Dynamics [81.80010043113445]
  Local weight fine-tuning, LoRA-based adaptation, and activation-based interventions are studied in isolation.<n>We present a unified view that frames these interventions as dynamic weight updates induced by a control signal.<n>Across methods, we observe a consistent trade-off between preference and utility: stronger control increases preference while predictably reducing utility.
  arXiv  Detail & Related papers  (2026-02-02T17:04:36Z)
• On the Paradoxical Interference between Instruction-Following and Task Solving [50.75960598434753]
  Instruction following aims to align Large Language Models (LLMs) with human intent by specifying explicit constraints on how tasks should be performed.<n>We reveal a counterintuitive phenomenon: instruction following can paradoxically interfere with LLMs' task-solving capability.<n>We propose a metric, SUSTAINSCORE, to quantify the interference of instruction following with task solving.
  arXiv  Detail & Related papers  (2026-01-29T17:48:56Z)
• Shared Parameter Subspaces and Cross-Task Linearity in Emergently Misaligned Behavior [9.202404525860189]
  We show that large language models can develop broadly misaligned behaviors after being fine-tuned on narrowly harmful datasets.<n>Our results indicate that EM arises from different narrow tasks discovering the same set of shared parameter directions.
  arXiv  Detail & Related papers  (2025-11-03T19:50:24Z)
• Sycophancy Is Not One Thing: Causal Separation of Sycophantic Behaviors in LLMs [1.9693252204587723]
  We decompose sycophantic agreement and sycophantic praise, contrasting both with genuine agreement.<n>Results suggest that sycophantic behaviors correspond to distinct, independently steerable representations.
  arXiv  Detail & Related papers  (2025-09-25T15:19:39Z)
• Re-Emergent Misalignment: How Narrow Fine-Tuning Erodes Safety Alignment in LLMs [0.0]
  We show that fine tuning on insecure code induces internal changes that oppose alignment.<n>We identify a shared latent dimension in the model's activation space that governs alignment behavior.
  arXiv  Detail & Related papers  (2025-07-04T15:36:58Z)
• Model Editing as a Double-Edged Sword: Steering Agent Ethical Behavior Toward Beneficence or Harm [57.00627691433355]
  We frame agent behavior steering as a model editing task, which we term Behavior Editing.<n>We introduce BehaviorBench, a benchmark grounded in psychological moral theories.<n>We demonstrate that Behavior Editing can be used to promote ethical and benevolent behavior or, conversely, to induce harmful or malicious behavior.
  arXiv  Detail & Related papers  (2025-06-25T16:51:51Z)
• REAL: Reading Out Transformer Activations for Precise Localization in Language Model Steering [26.428347164111926]
  Inference-time steering aims to alter a large language model's responses without changing its parameters.<n>Existing approaches often rely on simplistic cues or ad hoc generalizations.<n>We introduce REAL, a framework for identifying behavior-relevant modules in Transformer models.
  arXiv  Detail & Related papers  (2025-06-10T02:16:50Z)
• Interference and Generalization in Temporal Difference Learning [86.31598155056035]
  We study the link between generalization and interference in temporal-difference (TD) learning. We find that TD easily leads to low-interference, under-generalizing parameters, while the effect seems reversed in supervised learning.
  arXiv  Detail & Related papers  (2020-03-13T15:49:58Z)

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.