Weighted InfoNCE objectives realize specific target geometries in embedding space, with SupCon producing size-dependent inter-class similarities under imbalance while Soft SupCon and certain continuous variants preserve regular simplex or unique optima.
hub
International conference on machine learning , pages=
87 Pith papers cite this work. Polarity classification is still indexing.
87
Pith papers citing it
hub tools
citation-role summary
background 2
method 1
citation-polarity summary
claims ledger
- method For augmented samples with positive advantage ( ˆAi,t >0 ), If ϵexp → ∞ is permitted, monotonic policy improvement cannot be guaranteed. Proof.When the advantage is positive, the objective seeks to increase ρi,t(θ)>1 . For Standard Samples (mi = 0), the upper bound is1 +ϵ std. For Augmented Samples (mi = 1), the upper bound is relaxed to1 +ϵ exp (whereϵ exp > ϵ std): LCLIP aug (θ) = min(ρi,t(θ) ˆAi,t,(1 +ϵ exp) ˆAi,t).(21) By relaxing the upper bound, we allow the policy to take larger gradient
- background ¯Qmin(s,a) = min j∈{1,...,K} ¯Q¯θj(s,a),(10) which yields the backup operatorB π ¯Qmin. h-step (chunk) Bellman regression.Each critic is trained by regressing to the shared target: LTD(θi) =E (st,at,r(h) t ,st+h)∼D h Qθi(st,a t)− B π ¯Qmin(st,a t) 2i .(11) The correspondingh-step Bellman backup is Bπ ¯Qmin(st,a t) =r (h) t +γ h Ea′∼π(·|st+h) ¯Qmin(st+h,a ′) ,(12) where theh-step return is r(h) t = h−1X i=0 γi rt+i.(13) Cal-QL calibration regularizer.To enable a smooth transition from offlin
co-cited works
representative citing papers
M³Att poisons medical multimodal RAG by pairing covert textual misinformation with query-agnostic visual perturbations that increase retrieval of the bad content, causing LLMs to generate clinically plausible but incorrect responses.
Thermo-VL augments a frozen Molmo-7B VLM with a trainable thermal encoder and prompt-conditioned dual-attention fusion to improve cross-spectrum visual reasoning.
WikiVQABench is a human-curated collection of Wikipedia-based VQA items that require both visual evidence and external knowledge from Wikidata to answer correctly.
Linear-DPO replaces sigmoid utility with linear utility and adds EMA reference to improve preference alignment in diffusion and flow-matching text-to-image models.
Proposes weighted aggregation of clusters and self-distillation-driven token pruning to improve both accuracy and efficiency in ViT-based visual place recognition.
Object functionalization is cast as neural graph completion over a functional graph of parts, contacts, and motions, followed by geometry realization that also rectifies erroneous motions, demonstrated on furniture with a new paired dataset.
Introduces the task of counterfactual time series forecasting with textual conditions plus a text-attribution mechanism that improves accuracy by distinguishing mutable from immutable factors.
R-DMesh generates high-fidelity 4D meshes aligned to video by disentangling base mesh, motion, and a learned rectification jump offset inside a VAE, then using Triflow Attention and rectified-flow diffusion.
NAACA uses a neuro-inspired oscillatory working memory to gate attention in audio language models, raising AudioQwen's average precision from 53.5% to 70.6% on XD-Violence while cutting unnecessary calls.
DirectTryOn achieves state-of-the-art one-step virtual try-on performance by applying pure conditional transport, garment preservation loss, and self-consistency loss to straighten trajectories in pretrained generative models.
A text-supervised global layout embedding augments local patch representations in late-interaction VDR, yielding +2.4 nDCG@5 and +2.3 MAP@5 gains over ColPali/ColQwen baselines on ViDoRe-v2.
MaMi-HOI counters geometric forgetting in diffusion models via a Geometry-Aware Proximity Adapter for precise contacts and a Kinematic Harmony Adapter for natural whole-body postures in human-object interactions.
AstroAlertBench evaluates multimodal LLMs on astronomical classification accuracy, reasoning, and honesty using real ZTF alerts, revealing that high accuracy often diverges from self-assessed reasoning quality.
ProjLens shows that backdoor parameters in MLLMs are encoded in low-rank subspaces of the projector and that embeddings shift toward the target direction with magnitude linear in input norm, activating only on poisoned samples.
AnchorSeg uses ordered query banks of latent reasoning tokens plus a spatial anchor token and a Token-Mask Cycle Consistency loss to achieve 67.7% gIoU and 68.1% cIoU on the ReasonSeg benchmark.
PRISM lets pre-trained text-to-image models handle long prompts by breaking them into compositional parts, predicting noise separately, and merging outputs via energy-based conjunction, matching fine-tuned models while generalizing better to prompts over 500 tokens.
GaLa uses hypergraph representations of objects and a TriView encoder with contrastive learning to improve vision-language models on procedural planning benchmarks.
RoboDreamer factorizes video generation using language primitives to achieve compositional generalization in robot world models, outperforming monolithic baselines on unseen goals in RT-X.
Latent Consistency Models enable high-fidelity text-to-image generation in 2-4 steps by directly predicting solutions to the probability flow ODE in latent space, distilled from pre-trained LDMs.
EvoPrompt uses LLMs to run evolutionary operators on populations of prompts, outperforming human-engineered prompts by up to 25% on BIG-Bench Hard tasks across 31 datasets.
LLaMA-Adapter turns frozen LLaMA 7B into a capable instruction follower using only 1.2M new parameters and zero-init attention, matching Alpaca while extending to image-conditioned reasoning on ScienceQA and COCO.
SemiPrune uses a small labeled subset and semi-supervised pseudo-labeling to enable supervised dataset pruning methods, achieving state-of-the-art results on domain-specific, image-corrupted, and long-tailed datasets.
Introduces LOES, a constructive spectral method to select task-discriminative subspaces from intermediate layer embeddings, and GeoReg for enforcing simplicial class geometry during fine-tuning, with reported gains increasing with model depth across modalities.
citing papers explorer
-
A Unified Geometric Framework for Weighted Contrastive Learning
Weighted InfoNCE objectives realize specific target geometries in embedding space, with SupCon producing size-dependent inter-class similarities under imbalance while Soft SupCon and certain continuous variants preserve regular simplex or unique optima.
-
Knowledge Poisoning Attacks on Medical Multi-Modal Retrieval-Augmented Generation
M³Att poisons medical multimodal RAG by pairing covert textual misinformation with query-agnostic visual perturbations that increase retrieval of the bad content, causing LLMs to generate clinically plausible but incorrect responses.
-
Thermo-VL: Extending Vision-Language Models to Thermal Infrared Perception
Thermo-VL augments a frozen Molmo-7B VLM with a trainable thermal encoder and prompt-conditioned dual-attention fusion to improve cross-spectrum visual reasoning.
-
WikiVQABench: A Knowledge-Grounded Visual Question Answering Benchmark from Wikipedia and Wikidata
WikiVQABench is a human-curated collection of Wikipedia-based VQA items that require both visual evidence and external knowledge from Wikidata to answer correctly.
-
Linear-DPO: Linear Direct Preference Optimization for Diffusion and Flow-Matching Generative Models
Linear-DPO replaces sigmoid utility with linear utility and adds EMA reference to improve preference alignment in diffusion and flow-matching text-to-image models.
-
Faster or Stronger: Towards Flexible Visual Place Recognition via Weighted Aggregation and Token Pruning
Proposes weighted aggregation of clusters and self-distillation-driven token pruning to improve both accuracy and efficiency in ViT-based visual place recognition.
-
Functionalization via Structure Completion and Motion Rectification
Object functionalization is cast as neural graph completion over a functional graph of parts, contacts, and motions, followed by geometry realization that also rectifies erroneous motions, demonstrated on furniture with a new paired dataset.
-
What if Tomorrow is the World Cup Final? Counterfactual Time Series Forecasting with Textual Conditions
Introduces the task of counterfactual time series forecasting with textual conditions plus a text-attribution mechanism that improves accuracy by distinguishing mutable from immutable factors.
-
R-DMesh: Video-Guided 3D Animation via Rectified Dynamic Mesh Flow
R-DMesh generates high-fidelity 4D meshes aligned to video by disentangling base mesh, motion, and a learned rectification jump offset inside a VAE, then using Triflow Attention and rectified-flow diffusion.
-
NAACA: Training-Free NeuroAuditory Attentive Cognitive Architecture with Oscillatory Working Memory for Salience-Driven Attention Gating
NAACA uses a neuro-inspired oscillatory working memory to gate attention in audio language models, raising AudioQwen's average precision from 53.5% to 70.6% on XD-Violence while cutting unnecessary calls.
-
DirectTryOn: One-Step Virtual Try-On via Straightened Conditional Transport
DirectTryOn achieves state-of-the-art one-step virtual try-on performance by applying pure conditional transport, garment preservation loss, and self-consistency loss to straighten trajectories in pretrained generative models.
-
Beyond Bag-of-Patches: Learning Global Layout via Textual Supervision for Late-Interaction Visual Document Retrieval
A text-supervised global layout embedding augments local patch representations in late-interaction VDR, yielding +2.4 nDCG@5 and +2.3 MAP@5 gains over ColPali/ColQwen baselines on ViDoRe-v2.
-
MaMi-HOI: Harmonizing Global Kinematics and Local Geometry for Human-Object Interaction Generation
MaMi-HOI counters geometric forgetting in diffusion models via a Geometry-Aware Proximity Adapter for precise contacts and a Kinematic Harmony Adapter for natural whole-body postures in human-object interactions.
-
AstroAlertBench: Evaluating the Accuracy, Reasoning, and Honesty of Multimodal LLMs in Astronomical Classification
AstroAlertBench evaluates multimodal LLMs on astronomical classification accuracy, reasoning, and honesty using real ZTF alerts, revealing that high accuracy often diverges from self-assessed reasoning quality.
-
ProjLens: Unveiling the Role of Projectors in Multimodal Model Safety
ProjLens shows that backdoor parameters in MLLMs are encoded in low-rank subspaces of the projector and that embeddings shift toward the target direction with magnitude linear in input norm, activating only on poisoned samples.
-
AnchorSeg: Language Grounded Query Banks for Reasoning Segmentation
AnchorSeg uses ordered query banks of latent reasoning tokens plus a spatial anchor token and a Token-Mask Cycle Consistency loss to achieve 67.7% gIoU and 68.1% cIoU on the ReasonSeg benchmark.
-
Long-Text-to-Image Generation via Compositional Prompt Decomposition
PRISM lets pre-trained text-to-image models handle long prompts by breaking them into compositional parts, predicting noise separately, and merging outputs via energy-based conjunction, matching fine-tuned models while generalizing better to prompts over 500 tokens.
-
GaLa: Hypergraph-Guided Visual Language Models for Procedural Planning
GaLa uses hypergraph representations of objects and a TriView encoder with contrastive learning to improve vision-language models on procedural planning benchmarks.
-
RoboDreamer: Learning Compositional World Models for Robot Imagination
RoboDreamer factorizes video generation using language primitives to achieve compositional generalization in robot world models, outperforming monolithic baselines on unseen goals in RT-X.
-
Latent Consistency Models: Synthesizing High-Resolution Images with Few-Step Inference
Latent Consistency Models enable high-fidelity text-to-image generation in 2-4 steps by directly predicting solutions to the probability flow ODE in latent space, distilled from pre-trained LDMs.
-
EvoPrompt: Connecting LLMs with Evolutionary Algorithms Yields Powerful Prompt Optimizers
EvoPrompt uses LLMs to run evolutionary operators on populations of prompts, outperforming human-engineered prompts by up to 25% on BIG-Bench Hard tasks across 31 datasets.
-
LLaMA-Adapter: Efficient Fine-tuning of Language Models with Zero-init Attention
LLaMA-Adapter turns frozen LLaMA 7B into a capable instruction follower using only 1.2M new parameters and zero-init attention, matching Alpaca while extending to image-conditioned reasoning on ScienceQA and COCO.
-
Label-Efficient Dataset Pruning via Semi-Supervised Pseudo-Labeling
SemiPrune uses a small labeled subset and semi-supervised pseudo-labeling to enable supervised dataset pruning methods, achieving state-of-the-art results on domain-specific, image-corrupted, and long-tailed datasets.
-
Uncovering the Latent Potential of Deep Intermediate Representations
Introduces LOES, a constructive spectral method to select task-discriminative subspaces from intermediate layer embeddings, and GeoReg for enforcing simplicial class geometry during fine-tuning, with reported gains increasing with model depth across modalities.
-
STELLAR: Scaling 3D Perception Large Models for Autonomous Driving
STELLAR trains up to 500M-parameter multi-modal models on 50M driving scenes and reports empirical scaling trends plus new state-of-the-art results on the Waymo Open Dataset.
-
What Makes a Representation Good for Single-Cell Perturbation Prediction?
PerturbedVAE disentangles perturbation-specific signals from invariant gene expression structure to recover causal representations and improve out-of-distribution prediction in single-cell perturbation modeling.
-
Low Latency Gaze Tracking via Latent Optical Sensing
A hardware prototype performs gaze estimation by optically encoding task-relevant features with a microlens array and mask, captured on a 4x4 phototransistor array and decoded by a small neural network, reaching 3.4 ms latency with competitive accuracy.
-
Artificial Intolerance: Stigmatizing Language in Clinical Documentation Skews Large Language Model Decision-Making
Frontier LLMs exhibit bias from stigmatizing language in clinical vignettes across four conditions, skewing decisions toward less aggressive management, with limited mitigation from Chain-of-Thought or self-debiasing prompts.
-
Differentiable Optimization Layers for Guaranteed Fairness in Deep Learning
Introduces a fairness layer for deep learning models that guarantees output parity and an online primal-dual algorithm for aggregate fairness guarantees in streaming predictions with small batch sizes.
-
TTE-Flash: Accelerating Reasoning-based Multimodal Representations via Think-Then-Embed Tokens
TTE-Flash trains latent think tokens with CoT generation loss and embedding tokens with contrastive loss to deliver high-performance multimodal representations without generating explicit reasoning at inference time.
-
Letting the neural code speak: Automated characterization of monkey visual neurons through human language
Natural language descriptions generated via a closed-loop pipeline with digital twins capture the selectivity of most neurons in macaque V1 and V4, with synthesized images driving 96% of V4 neurons into the top or bottom 5% of natural-image response distributions.
-
Plan in Sandbox, Navigate in Open Worlds: Learning Physics-Grounded Abstracted Experience for Embodied Navigation
SAGE trains agents in physics-grounded semantic abstractions via RL with asymmetric clipping, achieving 53.21% LLM-Match Success on A-EQA (+9.7% over baseline) and encouraging physical robot transfer.
-
Post-hoc Selective Classification for Reliable Synthetic Image Detection
ReSIDe generalizes logit-based confidence scores to intermediate layers of synthetic image detectors and uses preference optimization to aggregate them, cutting area under the risk-coverage curve by up to 69.55% under covariate shifts.
-
Generative 3D Gaussians with Learned Density Control
DeG models 3D Gaussians via learned octree density and uses VecSeq Sobol re-indexing to turn set generation into sequence modeling, claiming SOTA quality in single-image-to-3D.
-
Escaping the Diversity Trap in Robotic Manipulation via Anchor-Centric Adaptation
Anchor-Centric Adaptation escapes the diversity trap by prioritizing repeated demonstrations at core anchors over broad coverage, yielding higher success rates under fixed data budgets in robotic manipulation.
-
Large Vision-Language Models Get Lost in Attention
In LVLMs, attention can be replaced by random Gaussian weights with little or no performance loss, indicating that current models get lost in attention rather than efficiently using visual context.
-
MUSE: Resolving Manifold Misalignment in Visual Tokenization via Topological Orthogonality
MUSE decouples reconstruction and semantic learning in visual tokenization via topological orthogonality, yielding SOTA generation quality and improved semantic performance over its teacher model.
-
Threshold-Guided Optimization for Visual Generative Models
A threshold-guided alignment method lets visual generative models be optimized directly from scalar human ratings instead of requiring paired preference data.
-
Interpreting V1 Population Activity via Image-Neural Latent Representation Alignment
DINA is a dual-tower contrastive model that aligns images with mouse V1 neural activity to enable decoding and shows that low-level visual structure, not semantics or fine details, primarily supports the alignment.
-
FACTOR: Counterfactual Training-Free Test-Time Adaptation for Open-Vocabulary Object Detection
FACTOR uses counterfactual image perturbations to quantify and suppress attribute-dependent predictions in open-vocabulary object detection, improving robustness on corrupted datasets without any training.
-
Injecting Distributional Awareness into MLLMs via Reinforcement Learning for Deep Imbalanced Regression
A plug-and-play RL method adds batch-level distributional supervision via CCC rewards to reduce regression-to-the-mean in MLLMs on imbalanced regression benchmarks.
-
TripoSG: High-Fidelity 3D Shape Synthesis using Large-Scale Rectified Flow Models
TripoSG generates high-fidelity 3D meshes from input images via a large-scale rectified flow transformer and hybrid-trained 3D VAE on a custom 2-million-sample dataset, claiming state-of-the-art fidelity and generalization.
-
VisionReward: Fine-Grained Multi-Dimensional Human Preference Learning for Image and Video Generation
VisionReward learns multi-dimensional human preferences for image and video generation via hierarchical assessment and linear weighting, outperforming VideoScore by 17.2% in prediction accuracy and yielding 31.6% higher win rates in text-to-video models.
-
Video Prediction Policy: A Generalist Robot Policy with Predictive Visual Representations
Video Prediction Policy conditions robot action learning on future-frame predictions inside fine-tuned video diffusion models, yielding 18.6% relative gains on Calvin ABC-D and 31.6% higher real-world success rates.
-
TraceVLA: Visual Trace Prompting Enhances Spatial-Temporal Awareness for Generalist Robotic Policies
Visual trace prompting improves spatial-temporal awareness in VLA models, delivering 10% gains on SimplerEnv and 3.5x on real-robot tasks.
-
CameraCtrl: Enabling Camera Control for Text-to-Video Generation
CameraCtrl enables accurate camera pose control in video diffusion models through a trained plug-and-play module and dataset choices emphasizing diverse camera trajectories with matching appearance.
-
Revisiting Feature Prediction for Learning Visual Representations from Video
V-JEPA models trained only on feature prediction from 2 million public videos achieve 81.9% on Kinetics-400, 72.2% on Something-Something-v2, and 77.9% on ImageNet-1K using frozen ViT-H/16 backbones.
-
LanguageBind: Extending Video-Language Pretraining to N-modality by Language-based Semantic Alignment
LanguageBind aligns video, infrared, depth, and audio to a frozen language encoder via contrastive learning on the new VIDAL-10M dataset, extending video-language pretraining to N modalities.
-
Demystifying CLIP Data
MetaCLIP curates balanced 400M-pair subsets from CommonCrawl that outperform CLIP data, reaching 70.8% zero-shot ImageNet accuracy on ViT-B versus CLIP's 68.3%.
-
Heterogeneity-Aware Dataset Scheduling for Efficient Audio Large Language Model Training
GST uses gradient-based affinity metrics to form dataset groups and applies progressive scheduling, achieving 30-40% faster convergence than uniform mixture training on 14 AudioQA datasets while matching or exceeding performance.