Skip to main content
arXiv is now an independent nonprofit! Learn more

Showing 1–12 of 12 results for author: Venhoff, C

Searching in archive cs. Search in all archives.
.
  1. arXiv:2608.09928  [pdf, ps, other

    cs.CV cs.AI cs.CL cs.LG

    Multimodal Model Diffing for Feature Discovery and Control

    Authors: Hunar Batra, Lachin Naghashyar, Ashkan Khakzar, Philip Torr, Christian Schroeder de Witt, Constantin Venhoff, Ronald Clark

    Abstract: Multimodal Large Language Models (MLLMs) exhibit strong visual understanding, yet the internal features that cause these behaviors remain difficult to identify, audit, or control. While applicable to post-hoc inspection, hidden states that are decomposed into interpretable feature directions using sparse autoencoders (SAEs) neither readily isolate which features are changed by multimodal training,… ▽ More

    Submitted 10 August, 2026; originally announced August 2026.

    Comments: Preprint. Accepted at ICML 2026 Trustworthy AI for Good Workshop

  2. arXiv:2606.24251  [pdf, ps, other

    cs.AI

    Probing the Misaligned Thinking Process of Language Models

    Authors: Kaiwen Zhou, Constantin Venhoff, Jonathan Michala, Xin Eric Wang, William Saunders

    Abstract: Large language models exhibit a growing range of misaligned behaviors such as strategic deception, sandbagging, and self-preservation. As they are increasingly deployed in high-stakes settings, it is critical to reliably detect such behaviors to ensure safe and responsible use. In this work, we propose to monitor misalignment by decomposing it into fine-grained cognitive processes -- misalignment… ▽ More

    Submitted 23 June, 2026; originally announced June 2026.

  3. arXiv:2602.08713  [pdf, ps, other

    cs.CV cs.LG

    Towards Understanding Multimodal Fine-Tuning: Spatial Features

    Authors: Lachin Naghashyar, Hunar Batra, Ashkan Khakzar, Philip Torr, Ronald Clark, Christian Schroeder de Witt, Constantin Venhoff

    Abstract: Contemporary Vision-Language Models (VLMs) achieve strong performance on a wide range of tasks by pairing a vision encoder with a pre-trained language model, fine-tuned for visual-text inputs. Yet despite these gains, it remains unclear how language backbone representations adapt during multimodal training and when vision-specific capabilities emerge. In this work, we present the first mechanistic… ▽ More

    Submitted 6 February, 2026; originally announced February 2026.

  4. arXiv:2601.20075  [pdf, ps, other

    cs.CV

    Sparse CLIP: Co-Optimizing Interpretability and Performance in Contrastive Learning

    Authors: Chuan Qin, Constantin Venhoff, Sonia Joseph, Fanyi Xiao, Stefan Scherer

    Abstract: Contrastive Language-Image Pre-training (CLIP) has become a cornerstone in vision-language representation learning, powering diverse downstream tasks and serving as the default vision backbone in multimodal large language models (MLLMs). Despite its success, CLIP's dense and opaque latent representations pose significant interpretability challenges. A common assumption is that interpretability and… ▽ More

    Submitted 27 January, 2026; originally announced January 2026.

  5. arXiv:2512.03276  [pdf, ps, other

    cs.LG

    Too Late to Recall: Explaining the Two-Hop Problem in Multimodal Knowledge Retrieval

    Authors: Constantin Venhoff, Ashkan Khakzar, Sonia Joseph, Philip Torr, Neel Nanda

    Abstract: Training vision language models (VLMs) aims to align visual representations from a vision encoder with the textual representations of a pretrained large language model (LLM). However, many VLMs exhibit reduced factual recall performance compared to their LLM backbones, raising the question of how effective multimodal fine-tuning is at extending existing mechanisms within the LLM to visual inputs.… ▽ More

    Submitted 2 December, 2025; originally announced December 2025.

  6. arXiv:2510.07364  [pdf, ps, other

    cs.AI cs.LG

    Base Models Know How to Reason, Thinking Models Learn When

    Authors: Constantin Venhoff, Iván Arcuschin, Philip Torr, Arthur Conmy, Neel Nanda

    Abstract: What do thinking language models learn during training that their base models lack? We first present an unsupervised method that discovers a model's reasoning behaviors by training small Sparse Autoencoders on sentence-level activations of reasoning traces, yielding interpretable reasoning taxonomies. Building on this, we introduce constructive model diffing, which aims to reconstruct the base-to-… ▽ More

    Submitted 7 July, 2026; v1 submitted 8 October, 2025; originally announced October 2025.

    Comments: Accepted as a Spotlight at the International Conference on Machine Learning 2026

  7. arXiv:2508.20570  [pdf, ps, other

    cs.CV cs.AI

    Dyslexify: A Mechanistic Defense Against Typographic Attacks in CLIP

    Authors: Lorenz Hufe, Constantin Venhoff, Erblina Purelku, Maximilian Dreyer, Sebastian Lapuschkin, Wojciech Samek

    Abstract: Typographic attacks exploit multi-modal systems by injecting text into images, leading to targeted misclassifications, malicious content generation and even Vision-Language Model jailbreaks. In this work, we analyze how CLIP vision encoders behave under typographic attacks, locating specialized attention heads in the latter half of the model's layers that causally extract and transmit typographic… ▽ More

    Submitted 26 February, 2026; v1 submitted 28 August, 2025; originally announced August 2025.

  8. arXiv:2507.12638  [pdf, ps, other

    cs.LG

    Reasoning-Finetuning Repurposes Latent Representations in Base Models

    Authors: Jake Ward, Chuqiao Lin, Constantin Venhoff, Neel Nanda

    Abstract: Backtracking, an emergent behavior elicited by reasoning fine-tuning, has been shown to be a key mechanism in reasoning models' enhanced capabilities. Prior work has succeeded in manipulating this behavior via steering vectors, but the underlying mechanism remains poorly understood. In this work, we show that the emergence of backtracking in DeepSeek-R1-Distill-Llama-8B is in part driven by a repu… ▽ More

    Submitted 16 July, 2025; originally announced July 2025.

    Comments: 6 pages, 6 figures. ICML 2025 Workshop on Actionable Interpretability

  9. arXiv:2506.18167  [pdf, ps, other

    cs.LG cs.AI

    Understanding Reasoning in Thinking Language Models via Steering Vectors

    Authors: Constantin Venhoff, Iván Arcuschin, Philip Torr, Arthur Conmy, Neel Nanda

    Abstract: Recent advances in large language models (LLMs) have led to the development of thinking language models that generate extensive internal reasoning chains before producing responses. While these models achieve improved performance, controlling their reasoning processes remains challenging. This work presents a steering approach for thinking LLMs by analyzing and manipulating specific reasoning beha… ▽ More

    Submitted 22 October, 2025; v1 submitted 22 June, 2025; originally announced June 2025.

    Comments: Accepted to the Workshop on Reasoning and Planning for Large Language Models at ICLR 2025

  10. arXiv:2506.11976  [pdf, ps, other

    cs.CV cs.LG

    How Visual Representations Map to Language Feature Space in Multimodal LLMs

    Authors: Constantin Venhoff, Ashkan Khakzar, Sonia Joseph, Philip Torr, Neel Nanda

    Abstract: Effective multimodal reasoning depends on the alignment of visual and linguistic representations, yet the mechanisms by which vision-language models (VLMs) achieve this alignment remain poorly understood. Following the LiMBeR framework, we deliberately maintain a frozen large language model (LLM) and a frozen vision transformer (ViT), connected solely by training a linear adapter during visual ins… ▽ More

    Submitted 21 June, 2025; v1 submitted 13 June, 2025; originally announced June 2025.

  11. arXiv:2503.07639  [pdf, other

    cs.LG cs.CL

    Mixture of Experts Made Intrinsically Interpretable

    Authors: Xingyi Yang, Constantin Venhoff, Ashkan Khakzar, Christian Schroeder de Witt, Puneet K. Dokania, Adel Bibi, Philip Torr

    Abstract: Neurons in large language models often exhibit \emph{polysemanticity}, simultaneously encoding multiple unrelated concepts and obscuring interpretability. Instead of relying on post-hoc methods, we present \textbf{MoE-X}, a Mixture-of-Experts (MoE) language model designed to be \emph{intrinsically} interpretable. Our approach is motivated by the observation that, in language models, wider networks… ▽ More

    Submitted 5 March, 2025; originally announced March 2025.

  12. arXiv:2410.07456  [pdf, other

    cs.LG

    SAGE: Scalable Ground Truth Evaluations for Large Sparse Autoencoders

    Authors: Constantin Venhoff, Anisoara Calinescu, Philip Torr, Christian Schroeder de Witt

    Abstract: A key challenge in interpretability is to decompose model activations into meaningful features. Sparse autoencoders (SAEs) have emerged as a promising tool for this task. However, a central problem in evaluating the quality of SAEs is the absence of ground truth features to serve as an evaluation gold standard. Current evaluation methods for SAEs are therefore confronted with a significant trade-o… ▽ More

    Submitted 9 October, 2024; originally announced October 2024.