Investigating task-specific prompts and sparse autoencoders for activation monitoring

• URL: http://arxiv.org/abs/2504.20271v1
• Date: Mon, 28 Apr 2025 21:28:17 GMT
• Title: Investigating task-specific prompts and sparse autoencoders for activation monitoring
• Authors: Henk Tillman, Dan Mossing,
• Abstract summary: Internal activations of language models encode additional information that could be useful for this.<n>Recent work has proposed several approaches which may improve on naive linear probing.<n>We develop and test novel refinements of these methods and compare them against each other.
• Score: 0.0
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Language models can behave in unexpected and unsafe ways, and so it is valuable to monitor their outputs. Internal activations of language models encode additional information that could be useful for this. The baseline approach for activation monitoring is some variation of linear probing on a particular layer: starting from a labeled dataset, train a logistic regression classifier on that layer's activations. Recent work has proposed several approaches which may improve on naive linear probing, by leveraging additional computation. One class of techniques, which we call "prompted probing," leverages test time computation to improve monitoring by (1) prompting the model with a description of the monitoring task, and (2) applying a learned linear probe to resulting activations. Another class of techniques uses computation at train time: training sparse autoencoders offline to identify an interpretable basis for the activations, and e.g. max-pooling activations across tokens using that basis before applying a linear probe. However, one can also prompt the model with a description of the monitoring task and use its output directly. We develop and test novel refinements of these methods and compare them against each other. We find asking the model zero-shot is a reasonable baseline when inference-time compute is not limited; however, activation probing methods can substantially outperform this baseline given sufficient training data. Specifically, we recommend prompted probing when inference-time compute is available, due to its superior data efficiency and good generalization performance. Alternatively, if inference-time compute is limited, we find SAE-based probing methods outperform raw activation probing.

Related papers

• Test-time Offline Reinforcement Learning on Goal-related Experience [50.94457794664909]
  Research in foundation models has shown that performance can be substantially improved through test-time training.<n>We propose a novel self-supervised data selection criterion, which selects transitions from an offline dataset according to their relevance to the current state.<n>Our goal-conditioned test-time training (GC-TTT) algorithm applies this routine in a receding-horizon fashion during evaluation, adapting the policy to the current trajectory as it is being rolled out.
  arXiv  Detail & Related papers  (2025-07-24T21:11:39Z)
• Is a Good Foundation Necessary for Efficient Reinforcement Learning? The Computational Role of the Base Model in Exploration [32.77845864484552]
  We introduce a new computational framework for RL with language models, in which the learner interacts with the model through a sampling oracle.<n>We show that coverage, while not necessary for data efficiency, lower bounds the runtime of any algorithm in our framework.<n>We introduce a new algorithm, SpannerSampling, which obtains optimal data efficiency and is computationally efficient whenever the pre-trained model enjoys sufficient coverage.
  arXiv  Detail & Related papers  (2025-03-10T15:31:42Z)
• Scaling Test-Time Compute Without Verification or RL is Suboptimal [70.28430200655919]
  We show that finetuning LLMs with verifier-based (VB) methods based on RL or search is far superior to verifier-free (VF) approaches based on distilling or cloning search traces, given a fixed amount of compute/data budget.<n>We corroborate our theory empirically on both didactic and math reasoning problems with 3/8B-sized pre-trained LLMs, where we find verification is crucial for scaling test-time compute.
  arXiv  Detail & Related papers  (2025-02-17T18:43:24Z)
• Scaling LLM Test-Time Compute Optimally can be More Effective than Scaling Model Parameters [27.656263126925815]
  We study the scaling of inference-time computation in LLMs. We find that in both cases, the effectiveness of different approaches to scaling test-time compute critically varies depending on the difficulty of the prompt.
  arXiv  Detail & Related papers  (2024-08-06T17:35:05Z)
• Adaptive Retention & Correction: Test-Time Training for Continual Learning [114.5656325514408]
  A common problem in continual learning is the classification layer's bias towards the most recent task. We name our approach Adaptive Retention & Correction (ARC) ARC achieves an average performance increase of 2.7% and 2.6% on the CIFAR-100 and Imagenet-R datasets.
  arXiv  Detail & Related papers  (2024-05-23T08:43:09Z)
• Batch Active Learning of Reward Functions from Human Preferences [33.39413552270375]
  Preference-based learning enables reliable labeling by querying users with preference questions. Active querying methods are commonly employed in preference-based learning to generate more informative data. We develop a set of novel algorithms that enable efficient learning of reward functions using as few data samples as possible.
  arXiv  Detail & Related papers  (2024-02-24T08:07:48Z)
• How Many Pretraining Tasks Are Needed for In-Context Learning of Linear Regression? [92.90857135952231]
  Transformers pretrained on diverse tasks exhibit remarkable in-context learning (ICL) capabilities. We study ICL in one of its simplest setups: pretraining a linearly parameterized single-layer linear attention model for linear regression.
  arXiv  Detail & Related papers  (2023-10-12T15:01:43Z)
• DST-Det: Simple Dynamic Self-Training for Open-Vocabulary Object Detection [72.25697820290502]
  This work introduces a straightforward and efficient strategy to identify potential novel classes through zero-shot classification. We refer to this approach as the self-training strategy, which enhances recall and accuracy for novel classes without requiring extra annotations, datasets, and re-training. Empirical evaluations on three datasets, including LVIS, V3Det, and COCO, demonstrate significant improvements over the baseline performance.
  arXiv  Detail & Related papers  (2023-10-02T17:52:24Z)
• Distributive Pre-Training of Generative Modeling Using Matrix-Product States [0.0]
  We consider an alternative training scheme utilizing basic tensor network operations, e.g., summation and compression. The training algorithm is based on compressing the superposition state constructed from all the training data in product state representation. We benchmark the algorithm on the MNIST dataset and show reasonable results for generating new images and classification tasks.
  arXiv  Detail & Related papers  (2023-06-26T15:46:08Z)
• Intersection of Parallels as an Early Stopping Criterion [64.8387564654474]
  We propose a method to spot an early stopping point in the training iterations without the need for a validation set. For a wide range of learning rates, our method, called Cosine-Distance Criterion (CDC), leads to better generalization on average than all the methods that we compare against.
  arXiv  Detail & Related papers  (2022-08-19T19:42:41Z)
• CAFA: Class-Aware Feature Alignment for Test-Time Adaptation [50.26963784271912]
  Test-time adaptation (TTA) aims to address this challenge by adapting a model to unlabeled data at test time. We propose a simple yet effective feature alignment loss, termed as Class-Aware Feature Alignment (CAFA), which simultaneously encourages a model to learn target representations in a class-discriminative manner.
  arXiv  Detail & Related papers  (2022-06-01T03:02:07Z)

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.