At the Edge of Understanding: Sparse Autoencoders Trace The Limits of Transformer Generalization

• URL: http://arxiv.org/abs/2606.26396v1
• Date: Wed, 24 Jun 2026 21:26:43 GMT
• Title: At the Edge of Understanding: Sparse Autoencoders Trace The Limits of Transformer Generalization
• Abstract summary: We present a mechanistic framework for delineating the precise contours of transformer model robustness.<n>We find that OOD inputs, including subtle typos and jailbreak prompts, drive language models to operate on an increased number of fallacious concepts in their internals.
• Score: 4.857288562417505
• License: http://creativecommons.org/licenses/by/4.0/
• Abstract: Pre-trained transformers have demonstrated remarkable generalization abilities, at times extending beyond the scope of their training data. Yet, real-world deployments often face unexpected or adversarial data that diverges from training data distributions. Without explicit mechanisms for handling such shifts, model reliability and safety degrade, urging more disciplined study of out-of-distribution (OOD) settings for transformers. By systematic experiments, we present a mechanistic framework for delineating the precise contours of transformer model robustness. We find that OOD inputs, including subtle typos and jailbreak prompts, drive language models to operate on an increased number of fallacious concepts in their internals. We leverage this device to quantify and understand the degree of distributional shift in prompts, enabling a mechanistically grounded fine-tuning strategy to robustify LLMs. Expanding the very notion of OOD from input data to a model's private computational processes, a new transformer diagnostic at inference time is a critical step toward making AI systems safe for deployment across science, business, and government.

Related papers

• From Noise to Narrative: Tracing the Origins of Hallucinations in Transformers [4.857288562417505]
  hallucinations in the output of a transformer model can be reliably predicted from the concept patterns embedded in transformer layer activations.<n>This collection of insights has immediate consequences for aligning AI models with human values, AI safety, opening the attack surface for potential adversarial attacks.
  arXiv  Detail & Related papers  (2025-09-08T17:50:45Z)
• How Out-of-Distribution Detection Learning Theory Enhances Transformer: Learnability and Reliability [10.056026416603006]
  This paper introduces the OOD detection Probably Approximately Correct (PAC) Theory for transformers.<n>It shows that outliers can be accurately represented and distinguished with sufficient data under conditions.<n>This approach yields a novel algorithm that ensures learnability and refines the decision boundaries between inliers and outliers.
  arXiv  Detail & Related papers  (2024-06-13T17:54:09Z)
• FaultFormer: Pretraining Transformers for Adaptable Bearing Fault Classification [7.136205674624813]
  We present a novel self-supervised pretraining and fine-tuning framework based on transformer models. In particular, we investigate different tokenization and data augmentation strategies to reach state-of-the-art accuracies. This introduces a new paradigm where models can be pretrained on unlabeled data from different bearings, faults, and machinery and quickly deployed to new, data-scarce applications.
  arXiv  Detail & Related papers  (2023-12-04T22:51:02Z)
• Transformers as Decision Makers: Provable In-Context Reinforcement Learning via Supervised Pretraining [25.669038513039357]
  This paper provides a theoretical framework that analyzes supervised pretraining for in-context reinforcement learning. We show transformers with ReLU attention can efficiently approximate near-optimal online reinforcement learning algorithms.
  arXiv  Detail & Related papers  (2023-10-12T17:55:02Z)
• SIP: Injecting a Structural Inductive Bias into a Seq2Seq Model by Simulation [75.14793516745374]
  We show how a structural inductive bias can be efficiently injected into a seq2seq model by pre-training it to simulate structural transformations on synthetic data. Our experiments show that our method imparts the desired inductive bias, resulting in better few-shot learning for FST-like tasks.
  arXiv  Detail & Related papers  (2023-10-01T21:19:12Z)
• Supervised Pretraining Can Learn In-Context Reinforcement Learning [96.62869749926415]
  In this paper, we study the in-context learning capabilities of transformers in decision-making problems. We introduce and study Decision-Pretrained Transformer (DPT), a supervised pretraining method where the transformer predicts an optimal action. We find that the pretrained transformer can be used to solve a range of RL problems in-context, exhibiting both exploration online and conservatism offline.
  arXiv  Detail & Related papers  (2023-06-26T17:58:50Z)
• Birth of a Transformer: A Memory Viewpoint [25.294093283819443]
  Large language models based on transformers have achieved great empirical successes. As they are deployed more widely, there is a growing need to better understand their internal mechanisms in order to make them more reliable. We study how transformers balance these two types of distributions of knowledge by considering a synthetic setup where tokens are generated from either global or context-specific bigrams.
  arXiv  Detail & Related papers  (2023-06-01T15:30:33Z)
• Emergent Agentic Transformer from Chain of Hindsight Experience [96.56164427726203]
  We show that a simple transformer-based model performs competitively with both temporal-difference and imitation-learning-based approaches. This is the first time that a simple transformer-based model performs competitively with both temporal-difference and imitation-learning-based approaches.
  arXiv  Detail & Related papers  (2023-05-26T00:43:02Z)
• Latent Positional Information is in the Self-Attention Variance of Transformer Language Models Without Positional Embeddings [68.61185138897312]
  We show that a frozen transformer language model encodes strong positional information through the shrinkage of self-attention variance. Our findings serve to justify the decision to discard positional embeddings and thus facilitate more efficient pretraining of transformer language models.
  arXiv  Detail & Related papers  (2023-05-23T01:03:40Z)
• Automatic Rule Induction for Efficient Semi-Supervised Learning [56.91428251227253]
  Semi-supervised learning has shown promise in allowing NLP models to generalize from small amounts of labeled data. Pretrained transformer models act as black-box correlation engines that are difficult to explain and sometimes behave unreliably. We propose tackling both of these challenges via Automatic Rule Induction (ARI), a simple and general-purpose framework.
  arXiv  Detail & Related papers  (2022-05-18T16:50:20Z)
• XAI for Transformers: Better Explanations through Conservative Propagation [60.67748036747221]
  We show that the gradient in a Transformer reflects the function only locally, and thus fails to reliably identify the contribution of input features to the prediction. Our proposal can be seen as a proper extension of the well-established LRP method to Transformers.
  arXiv  Detail & Related papers  (2022-02-15T10:47:11Z)

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.