Author name cluster

Yossi Gandelsman

Possible papers associated with this exact author name in Arrow. This page groups case-insensitive exact name matches and is not a full identity disambiguation profile.

13 papers

2 author rows

ICLR Conference 2025 Conference Paper

Interpreting and Editing Vision-Language Representations to Mitigate Hallucinations

Nick Jiang
Anish Kachinthaya
Suzanne Petryk
Yossi Gandelsman

We investigate the internal representations of vision-language models (VLMs) to address hallucinations, a persistent challenge despite advances in model size and training. We project VLMs’ internal image representations to their language vocabulary and observe more confident output probabilities on real objects than hallucinated objects. We additionally use these output probabilities to spatially localize real objects. Building on this approach, we introduce a knowledge erasure algorithm that removes hallucinations by linearly orthogonalizing image features with respect to hallucinated object features. We show that targeted edits to a model’s latent representations can reduce hallucinations by up to 25.7% on the COCO2014 dataset while preserving performance. Our findings demonstrate how a deeper understanding of VLMs’ latent representations can enhance reliability and enable novel capabilities, such as zero-shot segmentation.

Details

ICML Conference 2025 Conference Paper

Interpreting the Repeated Token Phenomenon in Large Language Models

Itay Yona
Ilia Shumailov
Jamie Hayes
Yossi Gandelsman

Large Language Models (LLMs), despite their impressive capabilities, often fail to accurately repeat a single word when prompted to, and instead output unrelated text. This unexplained failure mode represents a vulnerability, allowing even end users to diverge models away from their intended behavior. We aim to explain the causes for this phenomenon and link it to the concept of "attention sinks", an emergent LLM behavior crucial for fluency, in which the initial token receives disproportionately high attention scores. Our investigation identifies the neural circuit responsible for attention sinks and shows how long repetitions disrupt this circuit. We extend this finding to other nonrepeating sequences that exhibit similar circuit disruptions. To address this, we propose a targeted patch that effectively resolves the issue without negatively impacting the overall performance of the model. This study provides a mechanistic explanation for an LLM vulnerability, demonstrating how interpretability can diagnose and address issues, and offering insights that pave the way for more secure and reliable models.

Details

ICLR Conference 2025 Conference Paper

Interpreting the Second-Order Effects of Neurons in CLIP

Yossi Gandelsman
Alexei A. Efros
Jacob Steinhardt

We interpret the function of individual neurons in CLIP by automatically describing them using text. Analyzing the direct effects (i.e. the flow from a neuron through the residual stream to the output) or the indirect effects (overall contribution) fails to capture the neurons' function in CLIP. Therefore, we present the "second-order lens", analyzing the effect flowing from a neuron through the later attention heads, directly to the output. We find that these effects are highly selective: for each neuron, the effect is significant for <2% of the images. Moreover, each effect can be approximated by a single direction in the text-image space of CLIP. We describe neurons by decomposing these directions into sparse sets of text representations. The sets reveal polysemantic behavior - each neuron corresponds to multiple, often unrelated, concepts (e.g. ships and cars). Exploiting this neuron polysemy, we mass-produce "semantic" adversarial examples by generating images with concepts spuriously correlated to the incorrect class. Additionally, we use the second-order effects for zero-shot segmentation, outperforming previous methods. Our results indicate that a automated interpretation of neurons can be used for model deception and for introducing new model capabilities

Details

ICML Conference 2025 Conference Paper

LLMs can see and hear without any training

Kumar Ashutosh
Yossi Gandelsman
Xinlei Chen
Ishan Misra
Rohit Girdhar

We present MILS: Multimodal Iterative LLM Solver, a surprisingly simple, training-free approach, to imbue multimodal capabilities into your favorite LLM. Leveraging their innate ability to perform multi-step reasoning, MILS prompts the LLM to generate candidate outputs, each of which are scored and fed back iteratively, eventually generating a solution to the task. This enables various applications that typically require training specialized models on task-specific data. In particular, we establish a new state-of-the-art on emergent zero-shot image, video and audio captioning. MILS seamlessly applies to media generation as well, discovering prompt rewrites to improve text-to-image generation, and even edit prompts for style transfer! Finally, being a gradient-free optimization approach, MILS can invert multimodal embeddings into text, enabling applications like cross-modal arithmetic.

Details

NeurIPS Conference 2025 Conference Paper

Same Task, Different Circuits: Disentangling Modality-Specific Mechanisms in VLMs

Yaniv Nikankin
Dana Arad
Yossi Gandelsman
Yonatan Belinkov

Vision-Language models (VLMs) show impressive abilities to answer questions on visual inputs (e. g. , counting objects in an image), yet demonstrate higher accuracies when performing an analogous task on text (e. g. , counting words in a text). We investigate this accuracy gap by identifying and comparing the circuits---the task-specific computational sub-graphs---in different modalities. We show that while circuits are largely disjoint between modalities, they implement relatively similar functionalities: the differences lie primarily in processing modality-specific data positions (an image or a text sequence). Zooming in on the image data representations, we observe they become aligned with the higher-performing analogous textual representations only towards later layers, too late in processing to effectively influence subsequent positions. To overcome this, we patch the representations of visual data tokens from later layers back into earlier layers. In experiments with multiple tasks and models, this simple intervention closes a third of the performance gap between the modalities, on average. Our analysis sheds light on the multi-modal performance gap in VLMs and suggests a training-free approach for reducing it.

PDF Details

JMLR Journal 2025 Journal Article

Test-Time Training on Video Streams

Renhao Wang
Yu Sun
Arnuv Tandon
Yossi Gandelsman
Xinlei Chen
Alexei A. Efros
Xiaolong Wang

Prior work has established Test-Time Training (TTT) as a general framework to further improve a trained model at test time. Before making a prediction on each test instance, the model is first trained on the same instance using a self-supervised task such as reconstruction. We extend TTT to the streaming setting, where multiple test instances - video frames in our case - arrive in temporal order. Our extension is online TTT: The current model is initialized from the previous model, then trained on the current frame and a small window of frames immediately before. Online TTT significantly outperforms the fixed-model baseline for four tasks, on three real-world datasets. The improvements are more than 2.2x and 1.5x for instance and panoptic segmentation. Surprisingly, online TTT also outperforms its offline variant that accesses strictly more information, training on all frames from the entire test video regardless of temporal order. This finding challenges those in prior work using synthetic videos. We formalize a notion of locality as the advantage of online over offline TTT, and analyze its role with ablations and a theory based on bias-variance trade-off. [abs] [ pdf ][ bib ] [ code ] &copy JMLR 2025. ( edit, beta )

PDF Details

NeurIPS Conference 2025 Conference Paper

Vision Transformers Don't Need Trained Registers

Nicholas Jiang
Amil Dravid
Alexei Efros
Yossi Gandelsman

We investigate the mechanism underlying a previously identified phenomenon in Vision Transformers -- the emergence of high-norm tokens that lead to noisy attention maps (Darcet et al. , 2024). We observe that in multiple models (e. g. , CLIP, DINOv2), a sparse set of neurons is responsible for concentrating high-norm activations on outlier tokens, leading to irregular attention patterns and degrading downstream visual processing. While the existing solution for removing these outliers involves retraining models from scratch with additional learned $\textit{register tokens}$, we use our findings to create a training-free approach to mitigate these artifacts. By shifting the high-norm activations from our discovered $\textit{register neurons}$ into an additional untrained token, we can mimic the effect of register tokens on a model already trained without registers. We demonstrate that our method produces cleaner attention and feature maps, enhances performance over base models across multiple downstream visual tasks, and achieves results comparable to models explicitly trained with register tokens. We then extend test-time registers to off-the-shelf vision-language models, yielding cleaner attention-based, text-to-image attribution. Finally, we outline a simple mathematical model that reflects the observed behavior of register neurons and high norm tokens. Our results suggest that test-time registers effectively take on the role of register tokens at test-time, offering a training-free solution for any pre-trained model released without them.

PDF Details

ICLR Conference 2024 Conference Paper

Idempotent Generative Network

Assaf Shocher
Amil Dravid
Yossi Gandelsman
Inbar Mosseri
Michael Rubinstein
Alexei A. Efros

We propose a new approach for generative modeling based on training a neural network to be idempotent. An idempotent operator is one that can be applied sequentially without changing the result beyond the initial application, namely $f(f(z))=f(z)$. The proposed model $f$ is trained to map a source distribution (e.g, Gaussian noise) to a target distribution (e.g. realistic images) using the following objectives: (1) Instances from the target distribution should map to themselves, namely $f(x)=x$. We define the target manifold as the set of all instances that $f$ maps to themselves. (2) Instances that form the source distribution should map onto the defined target manifold. This is achieved by optimizing the idempotence term, $f(f(z))=f(z)$ which encourages the range of $f(z)$ to be on the target manifold. Under ideal assumptions such a process provably converges to the target distribution. This strategy results in a model capable of generating an output in one step, maintaining a consistent latent space, while also allowing sequential applications for refinement. Additionally, we find that by processing inputs from both target and source distributions, the model adeptly projects corrupted or modified data back to the target manifold. This work is a first step towards a ``global projector'' that enables projecting any input into a target data distribution.

Details

TMLR Journal 2024 Journal Article

IMProv: Inpainting-based Multimodal Prompting for Computer Vision Tasks

Jiarui Xu
Yossi Gandelsman
Amir Bar
Jianwei Yang
Jianfeng Gao
Trevor Darrell
Xiaolong Wang

In-context learning allows adapting a model to new tasks given a task description at test time. In this paper, we present IMProv - a generative model that is able to in-context learn visual tasks from multimodal prompts. Given a textual description of a visual task (e.g. “Left: input image, Right: foreground segmentation”), a few input-output visual examples, or both, the model in-context learns to solve it for a new test input. We train a masked generative transformer on a new dataset of figures from computer vision papers and their associated captions, together with a captioned large-scale image-text dataset. During inference time, we prompt the model with text and/or image task example(s) and have the model inpaint the corresponding output. We show that training our model with text conditioning and scaling the dataset size improves in-context learning for computer vision tasks by over $+10\%$ AP for Foreground Segmentation, over $+5\%$ gains in AP for Single Object Detection, and almost $20\%$ lower LPIPS in Colorization. Our emperical results suggest that vision and language prompts are complementary and it is advantageous to use both to achieve better in-context learning performance.

PDF Details

ICLR Conference 2024 Conference Paper

Interpreting CLIP's Image Representation via Text-Based Decomposition

Yossi Gandelsman
Alexei A. Efros
Jacob Steinhardt

We investigate the CLIP image encoder by analyzing how individual model components affect the final representation. We decompose the image representation as a sum across individual image patches, model layers, and attention heads, and use CLIP's text representation to interpret the summands. Interpreting the attention heads, we characterize each head's role by automatically finding text representations that span its output space, which reveals property-specific roles for many heads (e.g. location or shape). Next, interpreting the image patches, we uncover an emergent spatial localization within CLIP. Finally, we use this understanding to remove spurious features from CLIP and to create a strong zero-shot image segmenter. Our results indicate that scalable understanding of transformer models is attainable and can be used to repair and improve models.

Details

NeurIPS Conference 2024 Conference Paper

Interpreting the Weight Space of Customized Diffusion Models

Amil Dravid
Yossi Gandelsman
Kuan-Chieh Wang
Rameen Abdal
Gordon Wetzstein
Alexei A. Efros
Kfir Aberman

We investigate the space of weights spanned by a large collection of customized diffusion models. We populate this space by creating a dataset of over 60, 000 models, each of which is a base model fine-tuned to insert a different person's visual identity. We model the underlying manifold of these weights as a subspace, which we term $\textit{weights2weights}$. We demonstrate three immediate applications of this space that result in new diffusion models -- sampling, editing, and inversion. First, sampling a set of weights from this space results in a new model encoding a novel identity. Next, we find linear directions in this space corresponding to semantic edits of the identity (e. g. , adding a beard), resulting in a new model with the original identity edited. Finally, we show that inverting a single image into this space encodes a realistic identity into a model, even if the input image is out of distribution (e. g. , a painting). We further find that these linear properties of the diffusion model weight space extend to other visual concepts. Our results indicate that the weight space of fine-tuned diffusion models can behave as an interpretable $\textit{meta}$-latent space producing new models.

PDF Details DOI

NeurIPS Conference 2022 Conference Paper

Test-Time Training with Masked Autoencoders

Yossi Gandelsman
Yu Sun
Xinlei Chen
Alexei Efros

Test-time training adapts to a new test distribution on the fly by optimizing a model for each test input using self-supervision. In this paper, we use masked autoencoders for this one-sample learning problem. Empirically, our simple method improves generalization on many visual benchmarks for distribution shifts. Theoretically, we characterize this improvement in terms of the bias-variance trade-off.

PDF Details

NeurIPS Conference 2022 Conference Paper

Visual Prompting via Image Inpainting

Amir Bar
Yossi Gandelsman
Trevor Darrell
Amir Globerson
Alexei Efros

How does one adapt a pre-trained visual model to novel downstream tasks without task-specific finetuning or any model modification? Inspired by prompting in NLP, this paper investigates visual prompting: given input-output image example(s) of a new task at test time and a new input image, the goal is to automatically produce the output image, consistent with the given examples. We show that posing this problem as simple image inpainting -- literally just filling in a hole in a concatenated visual prompt image -- turns out to be surprisingly effective, provided that the inpainting algorithm has been trained on the right data. We train masked auto-encoders on a new dataset that we curated -- 88k unlabeled figures from academic papers sources on Arxiv. We apply visual prompting to these pretrained models and demonstrate results on various downstream image-to-image tasks, including foreground segmentation, single object detection, colorization, edge detection, etc. Project page: https: //yossigandelsman. github. io/visual_prompt

PDF Details