Ask or search anything...
Behind the Myth of Exploration in Policy Gradients
20 Aug 2025
In order to compute near-optimal policies with policy-gradient algorithms, it is common in practice to include intrinsic exploration terms in the learning objective. Although the effectiveness of these terms is usually justified by an intrinsic need to explore environments, we propose a novel analysis with the lens of numerical optimization. Two criteria are introduced on the learning objective and two others on its stochastic gradient estimates, and are afterwards used to discuss the quality of the policy after optimization. The analysis sheds light on two separate effects of exploration techniques. First, they make it possible to smooth the learning objective and to eliminate local optima while preserving the global maximum. Second, they modify the gradient estimates, increasing the probability that the stochastic parameter updates eventually provide an optimal policy. We empirically illustrate these effects with exploration strategies based on entropy bonuses, identifying limitations and suggesting directions for future work.
TrustLLM: Trustworthiness in Large Language Models
30 Sep 2024
Meng Jiang
Quanxin Mei
Tianyi Zhou
Michigan State University
University of Illinois at Urbana-ChampaignThe TRUSTLLM framework and benchmark offer a comprehensive system for evaluating the trustworthiness of large language models across six key dimensions. This work reveals that while proprietary models generally exhibit higher trustworthiness, open-source models can also achieve strong performance in specific areas, highlighting challenges like 'over-alignment' and data leakage.
View blog
PAGE-4D: Disentangled Pose and Geometry Estimation for VGGT-4D Perception
08 Dec 2025
Harvard University
Imperial College LondonRecent 3D feed-forward models, such as the Visual Geometry Grounded Transformer (VGGT), have shown strong capability in inferring 3D attributes of static scenes. However, since they are typically trained on static datasets, these models often struggle in real-world scenarios involving complex dynamic elements, such as moving humans or deformable objects like umbrellas. To address this limitation, we introduce PAGE-4D, a feedforward model that extends VGGT to dynamic scenes, enabling camera pose estimation, depth prediction, and point cloud reconstruction -- all without post-processing. A central challenge in multi-task 4D reconstruction is the inherent conflict between tasks: accurate camera pose estimation requires suppressing dynamic regions, while geometry reconstruction requires modeling them. To resolve this tension, we propose a dynamics-aware aggregator that disentangles static and dynamic information by predicting a dynamics-aware mask -- suppressing motion cues for pose estimation while amplifying them for geometry reconstruction. Extensive experiments show that PAGE-4D consistently outperforms the original VGGT in dynamic scenarios, achieving superior results in camera pose estimation, monocular and video depth estimation, and dense point map reconstruction.
Quantum computing and artificial intelligence: status and perspectives
30 Jun 2025
CNRSA collaborative white paper coordinated by the Quantum Community Network comprehensively analyzes the current status and future perspectives of Quantum Artificial Intelligence, categorizing its potential into "Quantum for AI" and "AI for Quantum" applications. It proposes a strategic research and development agenda to bolster Europe's competitive position in this rapidly converging technological domain.
View blog
Can outcome communication explain Bell nonlocality?
14 Oct 2025
University of WaterlooA central aspect of quantum information is that correlations between spacelike separated observers sharing entangled states cannot be reproduced by local hidden variable (LHV) models, a phenomenon known as Bell nonlocality. If one wishes to explain such correlations by classical means, a natural possibility is to allow communication between the parties. In particular, LHV models augmented with two bits of classical communication can explain the correlations of any two-qubit state. Would this still hold if communication is restricted to measurement outcomes? While in certain scenarios with a finite number of inputs the answer is yes, we prove that if a model must reproduce all projective measurements, then for any qubit-qudit state the answer is no. In fact, a qubit-qudit under projective measurements admits an LHV model with outcome communication if and only if it already admits an LHV model without communication. On the other hand, we also show that when restricted sets of measurements are considered (for instance, when the qubit measurements are in the upper hemisphere of the Bloch ball), outcome communication does offer an advantage. This exemplifies that trivial properties in standard LHV scenarios, such as deterministic measurements and outcome-relabelling, play a crucial role in the outcome communication scenario.
Explainability for Vision Foundation Models: A Survey
21 Jan 2025
As artificial intelligence systems become increasingly integrated into daily life, the field of explainability has gained significant attention. This trend is particularly driven by the complexity of modern AI models and their decision-making processes. The advent of foundation models, characterized by their extensive generalization capabilities and emergent uses, has further complicated this landscape. Foundation models occupy an ambiguous position in the explainability domain: their complexity makes them inherently challenging to interpret, yet they are increasingly leveraged as tools to construct explainable models. In this survey, we explore the intersection of foundation models and eXplainable AI (XAI) in the vision domain. We begin by compiling a comprehensive corpus of papers that bridge these fields. Next, we categorize these works based on their architectural characteristics. We then discuss the challenges faced by current research in integrating XAI within foundation models. Furthermore, we review common evaluation methodologies for these combined approaches. Finally, we present key observations and insights from our survey, offering directions for future research in this rapidly evolving field.
Knowledge distillation from multi-modal to mono-modal segmentation networks
Shanghai Jiao Tong UniversityA framework named KD-Net enables mono-modal medical image segmentation networks to leverage knowledge from multi-modal networks through generalized knowledge distillation. The approach boosts segmentation accuracy, achieving a Dice score of 71.67 for enhancing tumors on the BraTS 2018 dataset, outperforming a mono-modal baseline of 68.1.
View blog
Information-theoretic constraints in quantum gravity and cosmology
17 Oct 2025
In this dissertation, we review results on quantum information constraints in gravity that are relevant to cosmological models and demonstrate how this approach sheds light on cosmological holography. Using Jackiw-Teitelboim gravity as a toy model, we establish the validity of the quantum Bousso bound and prove the restricted quantum focusing conjecture (rQFC), which is a central assumption in semiclassical gravity and holography. Interestingly, we find violations of the original QFC in this model. Building on these results, we introduce a covariant prescription for computing holographic entanglement entropy in the static patch holography description of de Sitter space and explore a generalization to closed FLRW spacetimes. This leads us to a formulation of subregion-subregion duality, where entanglement between complementary holographic theories gives rise to bulk connectivity. Causality considerations imply that entanglement wedges are bounded by surfaces that are not extremal in the usual sense, but instead satisfy an extremization with constraint. Finally, we discuss the connected wedge theorem, which establishes a relation between the causal structure of spacetime and information-theoretic constraints in the holographic dual. We argue that maintaining the consistency of static patch holography with this theorem leads to constraints on the causal structure of the dual theory. Our results suggest a novel relationship between static patch holography and the dS/CFT correspondence.
Optimizing Backward Policies in GFlowNets via Trajectory Likelihood Maximization
03 Mar 2025
Generative Flow Networks (GFlowNets) are a family of generative models that
learn to sample objects with probabilities proportional to a given reward
function. The key concept behind GFlowNets is the use of two stochastic
policies: a forward policy, which incrementally constructs compositional
objects, and a backward policy, which sequentially deconstructs them. Recent
results show a close relationship between GFlowNet training and
entropy-regularized reinforcement learning (RL) problems with a particular
reward design. However, this connection applies only in the setting of a fixed
backward policy, which might be a significant limitation. As a remedy to this
problem, we introduce a simple backward policy optimization algorithm that
involves direct maximization of the value function in an entropy-regularized
Markov Decision Process (MDP) over intermediate rewards. We provide an
extensive experimental evaluation of the proposed approach across various
benchmarks in combination with both RL and GFlowNet algorithms and demonstrate
its faster convergence and mode discovery in complex environments.
Ignore the KL Penalty! Boosting Exploration on Critical Tokens to Enhance RL Fine-Tuning
10 Feb 2025
Université Paris-SaclayFrench researchers from Université Paris-Saclay and onepoint reveal fundamental relationships between pre-training depth and exploration capabilities in language models, introducing a novel "prioritized KL penalty" that significantly improves reinforcement learning fine-tuning efficiency by targeting critical decision points while maintaining model stability.
View blog
Fast reconstruction of degenerate populations of conductance-based neuron models from spike times
16 Sep 2025
Neurons communicate through spikes, and spike timing is a crucial part of neuronal processing. Spike times can be recorded experimentally both intracellularly and extracellularly, and are the main output of state-of-the-art neural probes. On the other hand, neuronal activity is controlled at the molecular level by the currents generated by many different transmembrane proteins called ion channels. Connecting spike timing to ion channel composition remains an arduous task to date. To address this challenge, we developed a method that combines deep learning with a theoretical tool called Dynamic Input Conductances (DICs), which reduce the complexity of ion channel interactions into three interpretable components describing how neurons spike. Our approach uses deep learning to infer DICs directly from spike times and then generates populations of "twin" neuron models that replicate the observed activity while capturing natural variability in membrane channel composition. The method is fast, accurate, and works using only spike recordings. We also provide open-source software with a graphical interface, making it accessible to researchers without programming expertise.
SplitFlow: Flow Decomposition for Inversion-Free Text-to-Image Editing
Rectified flow models have become a de facto standard in image generation due to their stable sampling trajectories and high-fidelity outputs. Despite their strong generative capabilities, they face critical limitations in image editing tasks: inaccurate inversion processes for mapping real images back into the latent space, and gradient entanglement issues during editing often result in outputs that do not faithfully reflect the target prompt. Recent efforts have attempted to directly map source and target distributions via ODE-based approaches without inversion; however,these methods still yield suboptimal editing quality. In this work, we propose a flow decomposition-and-aggregation framework built upon an inversion-free formulation to address these limitations. Specifically, we semantically decompose the target prompt into multiple sub-prompts, compute an independent flow for each, and aggregate them to form a unified editing trajectory. While we empirically observe that decomposing the original flow enhances diversity in the target space, generating semantically aligned outputs still requires consistent guidance toward the full target prompt. To this end, we design a projection and soft-aggregation mechanism for flow, inspired by gradient conflict resolution in multi-task learning. This approach adaptively weights the sub-target velocity fields, suppressing semantic redundancy while emphasizing distinct directions, thereby preserving both diversity and consistency in the final edited output. Experimental results demonstrate that our method outperforms existing zero-shot editing approaches in terms of semantic fidelity and attribute disentanglement. The code is available at this https URL.
TabStruct: Measuring Structural Fidelity of Tabular Data
University of CambridgeEvaluating tabular generators remains a challenging problem, as the unique causal structural prior of heterogeneous tabular data does not lend itself to intuitive human inspection. Recent work has introduced structural fidelity as a tabular-specific evaluation dimension to assess whether synthetic data complies with the causal structures of real data. However, existing benchmarks often neglect the interplay between structural fidelity and conventional evaluation dimensions, thus failing to provide a holistic understanding of model performance. Moreover, they are typically limited to toy datasets, as quantifying existing structural fidelity metrics requires access to ground-truth causal structures, which are rarely available for real-world datasets. In this paper, we propose a novel evaluation framework that jointly considers structural fidelity and conventional evaluation dimensions. We introduce a new evaluation metric, , which enables the assessment of structural fidelity even in the absence of ground-truth causal structures. In addition, we present , a comprehensive evaluation benchmark offering large-scale quantitative analysis on 13 tabular generators from nine distinct categories, across 29 datasets. Our results demonstrate that global utility provides a task-independent, domain-agnostic lens for tabular generator performance. We release the TabStruct benchmark suite, including all datasets, evaluation pipelines, and raw results. Code is available at this https URL.
Compression in 3D Gaussian Splatting: A Survey of Methods, Trends, and Future Directions
26 Feb 2025
A comprehensive survey systematically reviews and categorizes compression techniques for 3D Gaussian Splatting (3DGS), proposing a novel taxonomy of unstructured and structured methods to address the technology's substantial memory footprint. The analysis highlights that unstructured methods achieve significant FPS improvements (4x-7x faster) with good compression (up to 55x), while structured methods provide superior compression ratios (up to 188x) with comparable rendering speeds, charting a path for optimizing 3DGS for resource-constrained applications.
View blog
MATPAC++: Enhanced Masked Latent Prediction for Self-Supervised Audio Representation Learning
18 Aug 2025
Masked latent prediction has emerged as a leading paradigm in self-supervised learning (SSL), especially for general audio and music representation learning. While recent methods have demonstrated strong performance, the role of the predictor module used at the output of such SSL systems remains mainly overlooked, despite being crucial for solving the pretext task at hand. In particular, this module should be able to deal with the ambiguity inherent in audio content, especially when it is composed of multiple sound sources. This work proposes a novel enhancement: integrating Multiple Choice Learning (MCL) to explicitly model prediction ambiguity and improve representation quality. We build on top of the recently proposed MATPAC system, improving its prediction and unsupervised classification pretext tasks with MCL. We extensively evaluate our method, MATPAC++, through both linear probing across multiple downstream tasks and fine-tuning on AudioSet, employing a unified protocol that enables rigorous and fair comparisons with state-of-the-art SSL approaches. Results show that our proposal achieves state-of-the-art when fine-tuned on AudioSet and overall state-of-the-art scores on downstream tasks. Additionally, we examine domain specialisation by training exclusively on music data, where our model achieves state-of-the-art performance with significantly improved efficiency.
I-INR: Iterative Implicit Neural Representations
27 Nov 2025
배성호(인공지능학과)
Ali Haider
Researchers from Kyung Hee University, Adobe Research, Chung-Ang University, and Télécom Paris introduce Iterative Implicit Neural Representations (I-INRs), a plug-and-play framework that enhances existing Implicit Neural Representations (INRs) by incorporating an iterative refinement process. The method effectively mitigates spectral bias and improves noise robustness, achieving superior signal reconstruction quality across various tasks with minimal additional computational overhead.
View blog
First Order Motion Model for Image Animation
01 Oct 2020
Researchers at the University of Trento and Snap Inc. developed an object-agnostic framework for image animation that transfers motion from a driving video to a source image. The method introduces a first-order motion model that learns sparse keypoints alongside local affine transformations, enabling more accurate deformation modeling and producing highly realistic, coherent animations across diverse object categories, significantly outperforming prior methods.
View blog
Ask and Remember: A Questions-Only Replay Strategy for Continual Visual Question Answering
27 Jul 2025
InriaContinual Learning in Visual Question Answering (VQACL) requires models to acquire new visual-linguistic skills (plasticity) while preserving previously learned knowledge (stability). The inherent multimodality of VQACL exacerbates this challenge, as models must balance stability across visual and textual domains while adapting to novel objects and reasoning tasks. Existing methods, primarily designed for unimodal settings, often fall short in addressing this dual requirement. In this work, we present QUestion-only replay with Attention Distillation (QUAD), a novel approach for VQACL that leverages only past task questions for regularization. By eliminating the need to store visual data, QUAD not only reduces memory overhead, but also alleviates privacy concerns. Our method introduces a Question-only Replay mechanism that selectively reuses prior task questions to counteract overfitting to the answer space of the current task, addressing the problem out of answer set. Complementing this, we propose Attention Consistency Distillation to enforce both intra-modal and inter-modal attention consistency across tasks, preserving essential visual-linguistic associations. Extensive experiments on VQAv2 and NExT-QA demonstrate that QUAD significantly outperforms state-of-the-art methods, achieving robust performance in continual VQA. Code is available at: this https URL.
Revisiting Non-Acyclic GFlowNets in Discrete Environments
11 Sep 2025
Ян Максимов
Generative Flow Networks (GFlowNets) are a family of generative models that learn to sample objects from a given probability distribution, potentially known up to a normalizing constant. Instead of working in the object space, GFlowNets proceed by sampling trajectories in an appropriately constructed directed acyclic graph environment, greatly relying on the acyclicity of the graph. In our paper, we revisit the theory that relaxes the acyclicity assumption and present a simpler theoretical framework for non-acyclic GFlowNets in discrete environments. Moreover, we provide various novel theoretical insights related to training with fixed backward policies, the nature of flow functions, and connections between entropy-regularized RL and non-acyclic GFlowNets, which naturally generalize the respective concepts and theoretical results from the acyclic setting. In addition, we experimentally re-examine the concept of loss stability in non-acyclic GFlowNet training, as well as validate our own theoretical findings.
Table Foundation Models: on knowledge pre-training for tabular learning
New York UniversityTable foundation models bring high hopes to data science: pre-trained on tabular data to embark knowledge or priors, they should facilitate downstream tasks on tables. One specific challenge is that of data semantics: numerical entries take their meaning from context, e.g., column name. Pre-trained neural networks that jointly model column names and table entries have recently boosted prediction accuracy. While these models outline the promises of world knowledge to interpret table values, they lack the convenience of popular foundation models in text or vision. Indeed, they must be fine-tuned to bring benefits, come with sizeable computation costs, and cannot easily be reused or combined with other architectures. Here we introduce TARTE, a foundation model that transforms tables to knowledge-enhanced vector representations using the string to capture semantics. Pre-trained on large relational data, TARTE yields representations that facilitate subsequent learning with little additional cost. These representations can be fine-tuned or combined with other learners, giving models that push the state-of-the-art prediction performance and improve the prediction/computation performance trade-off. Specialized to a task or a domain, TARTE gives domain-specific representations that facilitate further learning. Our study demonstrates an effective approach to knowledge pre-training for tabular learning.
There are no more papers matching your filters at the moment.