Introspection Adapters: Training LLMs to Report Their Learned Behaviors

• URL: http://arxiv.org/abs/2604.16812v1
• Date: Sat, 18 Apr 2026 03:50:00 GMT
• Title: Introspection Adapters: Training LLMs to Report Their Learned Behaviors
• Abstract summary: We study a scalable approach to rapidly identify learned behaviors of many LLMs derived from a shared base LLM.<n>A single LoRA adapter jointly trained across the finetunes $M_i$ causes them to verbalize their implanted behaviors.<n>We find that this IA induces self-description of learned behaviors even in finetunes of $M$ that were trained in very different ways from the $M_i$.
• Score: 8.097590361201535
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: When model developers or users fine-tune an LLM, this can induce behaviors that are unexpected, deliberately harmful, or hard to detect. It would be far easier to audit LLMs if they could simply describe their behaviors in natural language. Here, we study a scalable approach to rapidly identify learned behaviors of many LLMs derived from a shared base LLM. Given a model $M$, our method works by finetuning models $M_i$ from $M$ with implanted behaviors $b_i$; the $(M_i, b_i)$ pairs serve as labeled training data. We then train an \emph{introspection adapter} (IA): a single LoRA adapter jointly trained across the finetunes $M_i$ to cause them to verbalize their implanted behaviors. We find that this IA induces self-description of learned behaviors even in finetunes of $M$ that were trained in very different ways from the $M_i$. For example, IAs generalize to AuditBench, achieving state-of-the-art at identifying explicitly hidden concerning behaviors. IAs can also be used to detect encrypted finetuning API attacks. They scale favorably with model size and training data diversity. Overall, our results suggest that IAs are a scalable, effective, and practically useful approach to auditing fine-tuned LLMs.

Related papers

• Learn-to-learn on Arbitrary Textual Conditioning: A Hypernetwork-Driven Meta-Gated LLM [5.325313097784103]
  In this paper, we activate the meta-signal of $$ within the SwiGLU blocks, resulting in a meta-gating mechanism that adaptively adjusts the nonlinearity of FFN.<n>By testing on different condition types such as task, domain, persona, and style, our method outperforms finetuning and meta-learning baselines.
  arXiv  Detail & Related papers  (2026-05-03T17:13:45Z)
• Where Did It Go Wrong? Attributing Undesirable LLM Behaviors via Representation Gradient Tracing [12.835224376066769]
  Large Language Models (LLMs) have demonstrated remarkable capabilities, yet their deployment is frequently undermined by undesirable behaviors.<n>We introduce a novel and efficient framework that diagnoses a range of undesirable LLM behaviors by analyzing representation and its gradients.<n>We systematically evaluate our method for tasks that include tracking harmful content, detecting backdoor poisoning, and identifying knowledge contamination.
  arXiv  Detail & Related papers  (2025-09-26T12:07:47Z)
• IF-GUIDE: Influence Function-Guided Detoxification of LLMs [53.051109450536885]
  We study how training data contributes to the emergence of toxic behaviors in large-language models.<n>We propose a $proactive approach that leverages influence functions to identify harmful tokens within any training data and suppress their impact during training.<n>We present a novel adaptation that measures token-level attributions from training data to model toxicity, along with techniques for selecting toxic training documents and a learning objective.
  arXiv  Detail & Related papers  (2025-06-02T15:32:36Z)
• Runaway is Ashamed, But Helpful: On the Early-Exit Behavior of Large Language Model-based Agents in Embodied Environments [54.67512489842682]
  Large language models (LLMs) have demonstrated strong planning and decision-making capabilities in complex embodied environments.<n>We take a first step toward exploring the early-exit behavior for LLM-based agents.
  arXiv  Detail & Related papers  (2025-05-23T08:23:36Z)
• Tell me about yourself: LLMs are aware of their learned behaviors [3.959641782135808]
  Behavioral self-awareness is relevant for AI safety, as models could use it to proactively disclose problematic behaviors.<n>Our results show that models have surprising capabilities for self-awareness and for the spontaneous articulation of implicit behaviors.
  arXiv  Detail & Related papers  (2025-01-19T17:28:12Z)
• Get my drift? Catching LLM Task Drift with Activation Deltas [55.75645403965326]
  Task drift allows attackers to exfiltrate data or influence the LLM's output for other users.<n>We show that a simple linear classifier can detect drift with near-perfect ROC AUC on an out-of-distribution test set.<n>We observe that this approach generalizes surprisingly well to unseen task domains, such as prompt injections, jailbreaks, and malicious instructions.
  arXiv  Detail & Related papers  (2024-06-02T16:53:21Z)
• Can Large Language Models Play Games? A Case Study of A Self-Play Approach [61.15761840203145]
  Large Language Models (LLMs) harness extensive data from the Internet, storing a broad spectrum of prior knowledge. Monte-Carlo Tree Search (MCTS) is a search algorithm that provides reliable decision-making solutions. This work introduces an innovative approach that bolsters LLMs with MCTS self-play to efficiently resolve turn-based zero-sum games.
  arXiv  Detail & Related papers  (2024-03-08T19:16:29Z)
• Do large language models and humans have similar behaviors in causal inference with script knowledge? [13.140513796801915]
  We study the processing of an event $B$ in a script-based story. In our manipulation, event $A$ is stated, negated, or omitted in an earlier section of the text.
  arXiv  Detail & Related papers  (2023-11-13T13:05:15Z)
• Making Harmful Behaviors Unlearnable for Large Language Models [50.44915524846857]
  Large language models (LLMs) have shown great potential as general-purpose AI assistants in various domains. LLMs can be easily fine-tuned into harmful assistants as the fine-tuning data often contains implicit or explicit harmful content. This paper proposes a controllable training framework that makes harmful behaviors unlearnable during the fine-tuning process.
  arXiv  Detail & Related papers  (2023-11-02T09:18:21Z)
• ReEval: Automatic Hallucination Evaluation for Retrieval-Augmented Large Language Models via Transferable Adversarial Attacks [91.55895047448249]
  This paper presents ReEval, an LLM-based framework using prompt chaining to perturb the original evidence for generating new test cases. We implement ReEval using ChatGPT and evaluate the resulting variants of two popular open-domain QA datasets. Our generated data is human-readable and useful to trigger hallucination in large language models.
  arXiv  Detail & Related papers  (2023-10-19T06:37:32Z)
• Understanding Emergent In-Context Learning from a Kernel Regression Perspective [55.95455089638838]
  Large language models (LLMs) have initiated a paradigm shift in transfer learning.<n>This paper proposes a kernel-regression perspective of understanding LLMs' ICL bahaviors when faced with in-context examples.<n>We find that during ICL, the attention and hidden features in LLMs match the behaviors of a kernel regression.
  arXiv  Detail & Related papers  (2023-05-22T06:45:02Z)

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.