论文检索

Deep Learning · Generative Models and Autoencoders

Wenqi Guo, Shan Du

We introduce Value Sign Flip (VSF), a simple and efficient method for incorporating negative prompt guidance in few-step (1-8 steps) diffusion and flow-matching image and video generation models. Unlike existing approaches such as classifier-free guidance (CFG), NASA, and NAG, VSF dynamically suppresses undesired content by flipping the sign of attention values from negative prompts. Our method requires only a small computational overhead and integrates effectively with MMDiT-style architectures such as Stable Diffusion 3.5 Turbo and Flux Schnell, as well as cross-attention-based models like Wan. We validate VSF on a proposed challenging dataset, NegGenBench, with complex prompt pairs. Experimental results on our proposed dataset show that VSF significantly improves negative prompt adherence (reaching 0.420 negative score for quality settings and 0.545 for strong settings) compared to prior methods in few-step models (scored 0.320-0.380 negative score) and even CFG in non-few-step models (scored 0.300 negative score), while maintaining competitive image quality and positive prompt adherence. Our method also suppressed a generate-then-edit pipeline, while also having a much faster runtime. Code, ComfyUI node, and dataset are available in https://github.com/weathon/VSF/tree/main.

Deep Learning · Generative Models and Autoencoders

Jiaming Fan, CAO DAMING, Xiangzhong Luo, Jiale Fu, CHONGHAN LIU, xu yang

Speculative Decoding (SD) is a key technique for accelerating Large Language Model (LLM) inference, but it typically requires training a draft model on a large dataset. We approach this problem from a data-centric perspective, finding that not all training samples contribute equally to the SD acceptance rate. Specifically, our theoretical analysis and empirical validation reveals that tokens inducing flatter predictive distributions from the target model are more valuable than those yielding sharply peaked distributions. Based on this insight, we propose flatness, a new metric to quantify this property, and develop the Sample-level-flatness-based Dataset Distillation (SFDD) approach, which filters the training data to retain only the most valuable samples. Experiments on the EAGLE framework demonstrate that SFDD can achieve over 2$\times$ training speedup using only 50\% of the data, while keeping the final model's inference speedup within 4\% of the full-dataset baseline. This work introduces an effective, data-centric approach that substantially improves the training efficiency for Speculative Decoding. Our code is available at https://github.com/fjm9933/Flatness.

Deep Learning · Generative Models and Autoencoders

Qingyuan Tian, Wenhong Zhu, Xiaoran Liu, Xiaofeng Wang, Rui Wang

Rotary Position Embedding (RoPE)-extension refers to modifying or generalizing the Rotary Position Embedding scheme to handle longer sequences than those encountered during pre-training. However, current extension strategies are highly diverse and lack a unified theoretical foundation. In this paper, we propose $\textbf{\textit{MrRoPE (Mixed-radix RoPE)}}$, a generalized encoding formulation based on a radix system conversion perspective, which elegantly unifies various RoPE-extension approaches as distinct radix conversion strategies. Based on this theory, we introduce two training-free extensions, $\textbf{\textit{MrRoPE-Uni}}$ and $\textbf{\textit{MrRoPE-Pro}}$, which leverage uniform and progressive radix conversion strategies, respectively, to achieve “train short, test long” generalization. Without fine-tuning, MrRoPE-Pro sustains over 85% recall in the 128K-context Needle-in-a-Haystack test and achieves more than double YaRN’s accuracy on Infinite-Bench retrieval and dialogue subsets. Theoretical analysis confirms that MrRoPE-Pro effectively raises the upper bound of RoPE's attainable encoding length, which further validates the reliability and utility of our theory and methodology.

Deep Learning · Generative Models and Autoencoders

Jing Jia, Sifan Liu, Bowen Song, Wei Yuan, Liyue Shen, Guanyang Wang

We systematically study antithetic initial noise in diffusion models, discovering that pairing each noise sample with its negation consistently produces strong negative correlation. This universal phenomenon holds across datasets, model architectures, conditional and unconditional sampling, and even other generative models such as VAEs and Normalizing Flows. To explain it, we combine experiments and theory and propose a \textit{symmetry conjecture} that the learned score function is approximately affine antisymmetric (odd symmetry up to a constant shift), supported by empirical evidence. This negative correlation leads to substantially more reliable uncertainty quantification with up to $90\%$ narrower confidence intervals. We demonstrate these gains on tasks including estimating pixel-wise statistics and evaluating diffusion inverse solvers. We also provide extensions with randomized quasi-Monte Carlo noise designs for uncertainty quantification, and explore additional applications of the antithetic noise design to improve image editing and generation diversity. Our framework is training-free, model-agnostic, and adds no runtime overhead. Code is available at https://github.com/jjia131/Antithetic-Noise-in-Diffusion-Models-page.

Deep Learning · Generative Models and Autoencoders

Amrith Setlur, Matthew Yang, Charlie Snell, Jeremiah Greer, Ian Wu, Virginia Smith, Max Simchowitz, Aviral Kumar

Test-time scaling offers a promising path to improve LLM reasoning by utilizing more compute at inference time; however, the true promise of this paradigm lies in extrapolation (i.e., improvement in performance on hard problems as LLMs keep "thinking" for longer, beyond the maximum token budget they were trained on). Surprisingly, we find that most existing reasoning models do not extrapolate well. We show that one way to enable extrapolation is by training the LLM to perform in-context exploration: training the LLM to effectively spend its test time budget by chaining operations (such as generation, verification, refinement, etc.), or testing multiple hypotheses before it commits to an answer. To enable in-context exploration, we identify three key ingredients as part of our recipe e3: (1) chaining skills that the base LLM has asymmetric competence in, e.g., chaining verification (easy) with generation (hard), as a way to implement in-context search; (2) leveraging "negative" gradients from incorrect traces to amplify exploration during RL, resulting in longer search traces that chains additional asymmetries; and (3) coupling task difficulty with training token budget during training via a specifically-designed curriculum to structure in-context exploration. Our recipe e3 produces the best known 1.7B model according to AIME'25 and HMMT'25 scores, and extrapolates to 2x the training token budget. Our e3-1.7B model not only attains high pass@1 scores, but also improves pass@k over the base model.

Deep Learning · Generative Models and Autoencoders

Uriel Singer, Yaron Lipman

Transition Matching (TM) is an emerging paradigm for generative modeling that generalizes diffusion and flow-matching models as well as continuous-state autoregressive models. TM, similar to previous paradigms, gradually transforms noise samples to data samples, however it uses a second ``internal'' generative model to implement the transition steps, making the transitions more expressive compared to diffusion and flow models. To make this paradigm tractable, TM employs a large backbone network and a smaller "head" module to efficiently execute the generative transition step. In this work, we present a large-scale, systematic investigation into the design, training and sampling of the head in TM frameworks, focusing on its time-continuous bidirectional variant. Through comprehensive ablations and experimentation involving training 56 different 1.7B text-to-image models (resulting in 549 unique evaluations) we evaluate the affect of the head module architecture and modeling during training as-well as a useful family of stochastic TM samplers. We analyze the impact on generation quality, training, and inference efficiency. We find that TM with an MLP head, trained with a particular time weighting and sampled with high frequency sampler provides best ranking across all metrics reaching state-of-the-art among all tested baselines, while Transformer head with sequence scaling and low frequency sampling is a runner up excelling at image aesthetics. Lastly, we believe the experiments presented highlight the design aspects that are likely to provide most quality and efficiency gains, while at the same time indicate what design choices are not likely to provide further gains.

General Machine Learning · Causality

Walter Nelson, Marco Fumero, THEOFANIS KARALETSOS, Francesco Locatello

Representation learning models exhibit a surprising stability in their internal representations. Whereas most prior work treats this stability as a single property, we formalize it as two distinct concepts: **statistical identifiability** (consistency of representations across runs) and **structural identifiability** (alignment of representations with some unobserved ground truth). Recognizing that perfect pointwise identifiability is generally unrealistic for modern representation learning models, we propose new model-agnostic definitions of statistical and structural near-identifiability of representations up to some error tolerance $\epsilon$. Leveraging these definitions, we prove a statistical $\epsilon$-**near-identifiability** result for the representations of models with nonlinear decoders, generalizing existing identifiability theory beyond last-layer representations in e.g. generative pre-trained transformers (GPTs) to near-identifiability of the intermediate representations of a broad class of models including (masked) autoencoders (MAEs) and supervised learners. Although these weaker assumptions confer weaker identifiability, we show that independent components analysis (ICA) can resolve much of the remaining linear ambiguity for this class of models, and validate and measure our near-identifiability claims empirically. With additional assumptions on the data-generating process, statistical identifiability extends to structural identifiability, yielding a simple and practical recipe for disentanglement: ICA post-processing of latent representations. On synthetic benchmarks, this approach achieves state-of-the-art disentanglement using a vanilla autoencoder. With a foundation model-scale MAE for cell microscopy, it disentangles biological variation from technical batch effects, substantially improving downstream generalization.

Deep Learning · Generative Models and Autoencoders

Tianyu Xie, Shuchen Xue, Zijin Feng, Tianyang Hu, Jiacheng Sun, Zhenguo Li, Cheng Zhang

Discrete diffusion models have recently shown great promise for modeling complex discrete data, with masked diffusion models (MDMs) offering a compelling trade-off between quality and generation speed. MDMs denoise by progressively unmasking multiple dimensions from an all-masked input, but their performance can degrade when using few denoising steps due to limited modeling of inter-dimensional dependencies. In this paper, we propose Variational Autoencoding Discrete Diffusion (VADD), a novel framework that enhances discrete diffusion with latent variable modeling to implicitly capture correlations among dimensions. By introducing an auxiliary recognition model, VADD enables stable training via variational lower bounds maximization and amortized inference over the training set. Our approach retains the efficiency of traditional MDMs while significantly improving sample quality, especially when the number of denoising steps is small. Empirical results on 2D toy data, pixel-level image generation, and text generation demonstrate that VADD consistently outperforms MDM baselines.

Theory · Interpretability and Visualization

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

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 performance are in tension: enforcing sparsity during training degrades accuracy, motivating recent post-hoc approaches such as Sparse Autoencoders (SAEs). However, these post-hoc approaches often suffer from degraded downstream performance and loss of CLIP's inherent multimodal capabilities, with most learned features remaining unimodal. We propose a simple yet effective approach that integrates sparsity directly into CLIP training, yielding representations that are both interpretable and performant. Compared to SAEs, our Sparse CLIP representations preserve strong downstream task performance, achieve superior interpretability, and retain multimodal capabilities. We show that multimodal sparse features enable straightforward semantic concept alignment and reveal training dynamics of how cross-modal knowledge emerges. Finally, as a proof of concept, we train a vision-language model on sparse CLIP representations that enables interpretable, vision-based steering capabilities. Our findings challenge conventional wisdom that interpretability requires sacrificing accuracy and demonstrate that interpretability and performance can be co-optimized, offering a promising design principle for future models.

Theory · Interpretability and Visualization

Usha Bhalla, Alex Oesterling, Claudio Mayrink Verdun, Hima Lakkaraju, Flavio Calmon

Translating the internal representations and computations of models into concepts that humans can understand is a key goal of interpretability. While recent dictionary learning methods such as Sparse Autoencoders (SAEs) provide a promising route to discover human-interpretable features, they often only recover token-specific, noisy, or highly local concepts. We argue that this limitation stems from neglecting the temporal structure of language, where semantic content typically evolves smoothly over sequences. Building on this insight, we introduce Temporal Sparse Autoencoders (T-SAEs), which incorporate a novel contrastive loss encouraging consistent activations of high-level features over adjacent tokens. This simple yet powerful modification enables SAEs to disentangle semantic from syntactic features in a self-supervised manner. Across multiple datasets and models, T-SAEs recover smoother, more coherent semantic concepts without sacrificing reconstruction quality. Strikingly, they exhibit clear semantic structure despite being trained without explicit semantic signal, offering a new pathway for unsupervised interpretability in language models.

Social Aspects · Accountability, Transparency and Interpretability

Rajiv Movva, Smitha Milli, Sewon Min, Emma Pierson

Human feedback can alter language models in unpredictable and undesirable ways, as practitioners lack a clear understanding of what feedback data encodes. While prior work studies preferences over certain attributes (e.g., length or sycophancy), automatically extracting relevant features without pre-specifying hypotheses remains challenging. We introduce *What's In My Human Feedback?* (WIMHF), a method to explain feedback data using sparse autoencoders. WIMHF characterizes both (1) the preferences a dataset is capable of measuring and (2) the preferences that the annotators actually express. Across 7 datasets, WIMHF identifies a small number of human-interpretable features that account for the majority of the preference prediction signal achieved by black-box models. These features reveal a wide diversity in what humans prefer, and the role of dataset-level context: for example, users on Reddit prefer informality and jokes, while annotators in HH-RLHF and PRISM disprefer them. WIMHF also surfaces potentially unsafe preferences, such as that LMArena users tend to vote against refusals, often in favor of toxic content. The learned features enable effective *data curation*: re-labeling the harmful examples in Arena yields large safety gains (+37%) with no cost to general performance. They also allow fine-grained *personalization*: on the Community Alignment dataset, we learn annotator-specific weights over subjective features that improve preference prediction. WIMHF provides a human-centered analysis method for practitioners to better understand and use preference data.

Social Aspects · Accountability, Transparency and Interpretability

Antonio De Santis, Schrasing Tong, Marco Brambilla, Lalana Kagal

Concept Bottleneck Models (CBMs) aim for ante-hoc interpretability by learning a bottleneck layer that predicts interpretable concepts before the decision. State-of-the-art approaches typically select which concepts to learn via human specification, open knowledge graphs, prompting an LLM, or using general CLIP concepts. However, concepts defined a-priori may not have sufficient predictive power for the task or even be learnable from the available data. As a result, these CBMs often significantly trail their black-box counterpart when controlling for information leakage. To address this, we introduce a novel CBM pipeline named Mechanistic CBM (M-CBM), which builds the bottleneck directly from a black-box model’s own learned concepts. These concepts are extracted via Sparse Autoencoders (SAEs) and subsequently named and annotated on a selected subset of images using a Multimodal LLM. For fair comparison and leakage control, we also introduce the Number of Contributing Concepts (NCC), a decision-level sparsity metric that extends the recently proposed NEC metric. Across diverse datasets, we show that M-CBMs consistently surpass prior CBMs at matched sparsity, while improving concept predictions and providing concise explanations. Our code is available at https://github.com/Antonio-Dee/M-CBM.

Applications · Health

Hiba Ahsan, Byron Wallace

LLMs are increasingly being used in healthcare. This promises to free physicians from drudgery, enabling better care to be delivered at scale. But the use of LLMs in this space also brings risks; for example, such models may worsen existing biases. How can we spot when LLMs are (spuriously) relying on patient race to inform predictions? In this work we assess the degree to which Sparse Autoencoders (SAEs) can reveal (and control) associations the model has made between race and stigmatizing concepts. We first identify SAE latents in gemma-2 models which appear to correlate with Black individuals. We find that this latent activates on reasonable input sequences (e.g., "African American") but also problematic words like "incarceration". We then show that we can use this latent to "steer" models to generate outputs about Black patients, and further that this can induce problematic associations in model outputs as a result. For example, activating the Black latent increases the risk assigned to the probability that a patient will become "belligerent". We also find that even in this controlled setting in which we causally intervene to manipulate only patient race, elicited CoT reasoning strings do not communicate that race is a factor in the resulting assessments. We evaluate the degree to which such "steering" via latents might be useful for mitigating bias. We find that this offers improvements in simple settings, but is less successful for more realistic and complex clinical tasks.

Computer Vision · Image and Video Generation

Boyang Zheng, Nanye Ma, Shengbang Tong, Saining Xie

Latent generative modeling has become the standard strategy for Diffusion Transformers (DiTs), but the autoencoder has barely evolved. Most DiTs still use the legacy VAE encoder, which introduces several limitations: large convolutional backbones that compromise architectural simplicity, low-dimensional latent spaces that restrict information capacity, and weak representations resulting from purely reconstruction-based training. In this work, we investigate replacing the VAE encoder–decoder with pretrained representation encoders (e.g., DINO, SigLIP, MAE) combined with trained decoders, forming what we call \emph{Representation Autoencoders} (RAEs). These models provide both high-quality reconstructions and semantically rich latent spaces, while allowing for a scalable transformer-based architecture. A key challenge arises in enabling diffusion transformers to operate effectively within these high-dimensional representations. We analyze the sources of this difficulty, propose theoretically motivated solutions, and validate them empirically. Our approach achieves faster convergence without auxiliary representation alignment losses. Using a DiT variant with a lightweight wide DDT-head, we demonstrate state-of-the-art image generation performance, reaching FIDs of 1.18 @256 resolution and 1.13 @512 on ImageNet.

Deep Learning · Generative Models and Autoencoders

Alex Ayoub, Kavosh Asadi, Dale Schuurmans, Csaba Szepesvari, Karim Bouyarmane

Large reasoning models (LRMs) often consume excessive tokens, inflating computational cost and latency. More broadly, in goal reaching sequential decision problems we often want to reach the goal quickly, and LRM reasoning can be viewed through this lens. We challenge the assumption that longer responses improve accuracy. By penalizing reasoning tokens using a discounted reinforcement learning setup (interpretable as a small token cost) and analyzing Blackwell optimality in restricted policy classes, we encourage concise yet accurate reasoning, analogous to preferring shorter successful trajectories in a stochastic shortest path problem. Experiments confirm our theoretical results that this approach shortens chains of thought while preserving accuracy.

Yanbo Chen, Weiwei Liu

Anomaly detection refers to the techniques that identify (probably unseen) rare or suspicious data that deviate significantly from the pre-defined normal data (Chalapathy & Chawla, 2019; Ruff et al., 2021). Empirical studies have observed that generative models trained on normal data tend to produce larger reconstruction errors when reconstructing anomalies. Based on this observation, researchers have developed various anomaly detection methods, referred to as reconstruction-based anomaly detection (RBAD) (Lv et al., 2024; Li et al., 2024) in the literature. Despite the empirical success of RBAD, the theoretical understanding of RBAD is still limited. This paper provides a theoretical analysis of RBAD. We analyze the training dynamics of a 2-layer convolutional autoencoder and introduce the cone set of the features. We prove that the cone sets of the normal features would absorb the (convolutional) kernels of the autoencoder during training and use these absorbed kernels to reconstruct the inputs. The absorbed kernels are more aligned with the normal features, which explains the cause of the reconstruction error gap between the normal data and the anomalies. Synthesized experiments are provided to validate our theoretical findings. We also visualize the training dynamics of the autoencoder on real-world data, demonstrating our proposed cone set intuition.

Deep Learning · Generative Models and Autoencoders

Ben Finkelshtein, Silviu Cucerzan, Sujay Kumar Jauhar, Ryen White

Large language models (LLMs) are increasingly leveraged for text-rich graph machine learning tasks, with node classification standing out due to its high-impact application domains such as fraud detection and recommendation systems. Yet, despite a surge of interest, the field lacks a principled understanding of the capabilities of LLMs in processing graph data. In this work, we conduct a large-scale, controlled evaluation across the key axes of variability: the LLM-graph interaction mode, comparing prompting, tool-use, and code generation; dataset domains, spanning citation, web-link, e-commerce, and social networks; homophilic vs. heterophilic regimes; short- vs. long-text features; LLM sizes and reasoning capabilities. We further analyze dependencies by independently truncating features, deleting edges, and removing labels to quantify reliance on input types. Our findings provide actionable guidance for both research and practice. (1) Code generation mode achieves the strongest overall performance, with especially large gains on long-text or high-degree graphs where prompting quickly exceeds the token budget. (2) All interaction strategies remain effective on heterophilic graphs, challenging the assumption that LLM-based methods collapse under low homophily. (3) Code generation mode is able to flexibly shift its reliance to the most informative input type, whether that be structure, features, or labels. Together, these results establish a clear picture of the strengths and limitations of current LLM–graph interaction modes and point to design principles for future methods.

Deep Learning · Generative Models and Autoencoders

Yanbo Dai, Zhenlan Ji, Zongjie Li, Shuai Wang

Model editing techniques are essential for efficiently updating knowledge in large language models (LLMs). However, the effectiveness of existing approaches degrades in massive editing scenarios, particularly when evaluated with practical metrics. Their robustness is also limited in context-rich settings or when editing multiple facts of the same subject simultaneously. We attribute these failures to the embedding misalignment among knowledge items, which undermines editing reliability at scale. To address this, we propose EAMET (Embedding Alignment Model Editing in Transformers), which addresses this issue by aligning the space of key and residual embeddings. Extensive experiments across six LLMs and three datasets demonstrate that EAMET consistently outperforms existing methods, achieving about 90\% editing efficacy when editing 10k facts.

Deep Learning · Generative Models and Autoencoders

Bowei He, Lihao Yin, Huiling Zhen, Xiaokun Zhang, Mingxuan Yuan, Chen Ma

Model pruning is an effective approach for compressing large language models (LLMs). However, this process often leads to significant degradation of model capabilities. While post-training techniques such as instruction tuning are commonly employed to recover model performance, existing methods often overlook the uneven deterioration of model capabilities and incur high computational costs. Moreover, some irrelevant instructions may also introduce negative effects to model capacity recovery. To address these challenges, we propose the **P**ost-training d**A**ta **S**election method for **E**fficient pruned large language model **R**ecovery (**PASER**). PASER aims to identify instructions to recover the most compromised model capacities with a certain data budget. Our approach first applies manifold learning and spectral clustering to group recovery instructions in the semantic space, revealing capability-specific instruction sets. Then, the data budget is adaptively allocated across clusters by the degree of corresponding model capability degradation. In each cluster, we prioritize data samples that lead to the most decline of model performance. To mitigate potential negative tuning effects, we also detect and filter out conflicting or irrelevant recovery data. Extensive experiments demonstrate that PASER significantly outperforms conventional baselines, effectively recovering the general capabilities of pruned LLMs while utilizing merely 4\%-20\% of the original post-training data. We provide the anonymous code repository in [Link](https://anonymous.4open.science/r/PASER-E606).

Computer Vision · 3D Rendering & Reconstruction

Binzhu Xie, Shi Qiu, Sicheng Zhang, Yinqiao Wang, Hao Xu, Muzammal Naseer, Chi-Wing Fu, Pheng-Ann Heng

Robust 3D hand reconstruction is challenging in egocentric vision due to depth ambiguity, self-occlusion, and complex hand-object interactions. Prior works attempt to mitigate the challenges by scaling up training data or incorporating auxiliary cues, often falling short of effectively handling unseen contexts. In this paper, we introduce EgoHandICL, the first in-context learning (ICL) framework for 3D hand reconstruction that achieves strong semantic alignment, visual consistency, and robustness under challenging egocentric conditions. Specifically, we develop (i) complementary exemplar retrieval strategies guided by vision–language models (VLMs), (ii) an ICL-tailored tokenizer that integrates multimodal context, and (iii) a Masked Autoencoders (MAE)-based architecture trained with 3D hand–guided geometric and perceptual objectives. By conducting comprehensive experiments on the ARCTIC and EgoExo4D benchmarks, our EgoHandICL consistently demonstrates significant improvements over state-of-the-art 3D hand reconstruction methods. We further show EgoHandICL’s applicability by testing it on real-world egocentric cases and integrating it with EgoVLMs to enhance their hand–object interaction reasoning. Our code and data will be publicly available.