Activation Oracles: Training and Evaluating LLMs as General-Purpose Activation Explainers

• URL: http://arxiv.org/abs/2512.15674v1
• Date: Wed, 17 Dec 2025 18:26:28 GMT
• Title: Activation Oracles: Training and Evaluating LLMs as General-Purpose Activation Explainers
• Authors: Adam Karvonen, James Chua, Clément Dumas, Kit Fraser-Taliente, Subhash Kantamneni, Julian Minder, Euan Ong, Arnab Sen Sharma, Daniel Wen, Owain Evans, Samuel Marks,
• Abstract summary: Large language model (LLM) activations are notoriously difficult to understand.<n>Recent work has proposed a simpler approach known as LatentQA.<n>We evaluate LatentQA-trained models in far out-of-distribution settings.
• Score: 10.70026676953023
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Large language model (LLM) activations are notoriously difficult to understand, with most existing techniques using complex, specialized methods for interpreting them. Recent work has proposed a simpler approach known as LatentQA: training LLMs to directly accept LLM activations as inputs and answer arbitrary questions about them in natural language. However, prior work has focused on narrow task settings for both training and evaluation. In this paper, we instead take a generalist perspective. We evaluate LatentQA-trained models, which we call Activation Oracles (AOs), in far out-of-distribution settings and examine how performance scales with training data diversity. We find that AOs can recover information fine-tuned into a model (e.g., biographical knowledge or malign propensities) that does not appear in the input text, despite never being trained with activations from a fine-tuned model. Our main evaluations are four downstream tasks where we can compare to prior white- and black-box techniques. We find that even narrowly-trained LatentQA models can generalize well, and that adding additional training datasets (such as classification tasks and a self-supervised context prediction task) yields consistent further improvements. Overall, our best AOs match or exceed prior white-box baselines on all four tasks and are the best method on 3 out of 4. These results suggest that diversified training to answer natural-language queries imparts a general capability to verbalize information about LLM activations.

Related papers

• Tiny language models [0.0]
  This study explores whether tiny language models (TLMs) exhibit the same key qualitative features of large language models (LLMs)<n>We demonstrate that TLMs exhibit a clear performance gap between pre-trained and non-pre-trained models across classification tasks.<n>The classification accuracy achieved by a pre-trained deep TLM architecture can be replicated through a soft committee of multiple, independently pre-trained shallow architectures.
  arXiv  Detail & Related papers  (2025-07-20T08:49:57Z)
• How Post-Training Reshapes LLMs: A Mechanistic View on Knowledge, Truthfulness, Refusal, and Confidence [52.9442657690445]
  Post-training is essential for the success of large language models (LLMs)<n>We compare base and post-trained LLMs from four perspectives to better understand post-training effects.
  arXiv  Detail & Related papers  (2025-04-03T06:30:55Z)
• SkillAggregation: Reference-free LLM-Dependent Aggregation [14.46141987797362]
  Large Language Models (LLMs) are increasingly used to assess NLP tasks. Recent work suggests using multiple LLMs as judges yields improved performance. This work focuses on aggregating predictions from multiple systems where no reference labels are available.
  arXiv  Detail & Related papers  (2024-10-14T07:13:47Z)
• LBC: Language-Based-Classifier for Out-Of-Variable Generalization [14.033963471962823]
  Large Language Models (LLMs) have great success in natural language processing tasks such as response generation. We find that the pre-trained knowledge of LLMs enables them to interpret new variables that appear in a test without additional training. We propose a Language-Based-Classifier (LBC) to maximize the benefits of LLMs to outperform TMLs on Out-of-Variable tasks.
  arXiv  Detail & Related papers  (2024-08-20T15:05:02Z)
• Generalization v.s. Memorization: Tracing Language Models' Capabilities Back to Pretraining Data [76.90128359866462]
  We introduce an extended concept of memorization, distributional memorization, which measures the correlation between the output probabilities and the pretraining data frequency.<n>This study demonstrates that memorization plays a larger role in simpler, knowledge-intensive tasks, while generalization is the key for harder, reasoning-based tasks.
  arXiv  Detail & Related papers  (2024-07-20T21:24:40Z)
• SELF-GUIDE: Better Task-Specific Instruction Following via Self-Synthetic Finetuning [70.21358720599821]
  Large language models (LLMs) hold the promise of solving diverse tasks when provided with appropriate natural language prompts. We propose SELF-GUIDE, a multi-stage mechanism in which we synthesize task-specific input-output pairs from the student LLM. We report an absolute improvement of approximately 15% for classification tasks and 18% for generation tasks in the benchmark's metrics.
  arXiv  Detail & Related papers  (2024-07-16T04:41:58Z)
• FRACTAL: Fine-Grained Scoring from Aggregate Text Labels [17.052047103156372]
  Large language models (LLMs) are increasingly tuned to power complex generation tasks such as writing, fact-seeking, querying and reasoning. Traditionally, human or model feedback for evaluating and tuning LLM performance has been provided at the response level. Recent works indicate that sentence-level labels may provide more accurate and interpretable feedback for LLM optimization.
  arXiv  Detail & Related papers  (2024-04-07T05:54:28Z)
• Learning to Generate Explainable Stock Predictions using Self-Reflective Large Language Models [54.21695754082441]
  We propose a framework to teach Large Language Models (LLMs) to generate explainable stock predictions. A reflective agent learns how to explain past stock movements through self-reasoning, while the PPO trainer trains the model to generate the most likely explanations. Our framework can outperform both traditional deep-learning and LLM methods in prediction accuracy and Matthews correlation coefficient.
  arXiv  Detail & Related papers  (2024-02-06T03:18:58Z)
• Supervised Knowledge Makes Large Language Models Better In-context Learners [94.89301696512776]
  Large Language Models (LLMs) exhibit emerging in-context learning abilities through prompt engineering. The challenge of improving the generalizability and factuality of LLMs in natural language understanding and question answering remains under-explored. We propose a framework that enhances the reliability of LLMs as it: 1) generalizes out-of-distribution data, 2) elucidates how LLMs benefit from discriminative models, and 3) minimizes hallucinations in generative tasks.
  arXiv  Detail & Related papers  (2023-12-26T07:24:46Z)
• LLM-augmented Preference Learning from Natural Language [19.700169351688768]
  Large Language Models (LLMs) are equipped to deal with larger context lengths. LLMs can consistently outperform the SotA when the target text is large. Few-shot learning yields better performance than zero-shot learning.
  arXiv  Detail & Related papers  (2023-10-12T17:17:27Z)
• TRACE: A Comprehensive Benchmark for Continual Learning in Large Language Models [52.734140807634624]
  Aligned large language models (LLMs) demonstrate exceptional capabilities in task-solving, following instructions, and ensuring safety. Existing continual learning benchmarks lack sufficient challenge for leading aligned LLMs. We introduce TRACE, a novel benchmark designed to evaluate continual learning in LLMs.
  arXiv  Detail & Related papers  (2023-10-10T16:38:49Z)
• Task Residual for Tuning Vision-Language Models [69.22958802711017]
  We propose a new efficient tuning approach for vision-language models (VLMs) named Task Residual Tuning (TaskRes) TaskRes explicitly decouples the prior knowledge of the pre-trained models and new knowledge regarding a target task. The proposed TaskRes is simple yet effective, which significantly outperforms previous methods on 11 benchmark datasets.
  arXiv  Detail & Related papers  (2022-11-18T15:09:03Z)

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.