ICLR 2024 Papers — Page 4
International Conference on Learning Representations · 2260 papers
CausalLM is not optimal for in-context learning
Nan Ding (Google Research), Radu Soricut (Google Research)
TransformerLarge Language ModelText
🎯 What it does: This paper conducts a theoretical analysis of the convergence properties of two commonly used architectures in Transformers—prefix language model (prefixLM) and causal language model (causalLM) in in-context learning (ICL), and verifies the performance differences between the two through multi-layer linear self-attention (LSA) and actual Softmax Transformer experiments.
Causally Aligned Curriculum Learning
Mingxuan Li (Columbia University), Elias Bareinboim (Columbia University)
OptimizationRobotic IntelligenceReinforcement LearningImage
🎯 What it does: This paper addresses the 'curse of dimensionality' problem in reinforcement learning by proposing a method to determine whether the source task and target task are aligned through causal graphs, thereby automatically generating an aligned curriculum learning sequence to enhance the learning efficiency of agents in environments with unobserved confounding factors.
CausalTime: Realistically Generated Time-series for Benchmarking of Causal Discovery
Yuxiao Cheng (Tsinghua University), Kunlun He (Chinese PLA General Hospital)
GenerationData SynthesisFlow-based ModelTime SeriesBenchmark
🎯 What it does: CausalTime proposes a complete process that starts from real time series data, fits dynamics through deep networks and regularization flows, extracts pseudo-causal graphs, splits them into causal, residual, and noise components, and then autoregressively generates new time series, providing the corresponding true causal graphs, thus forming a benchmark dataset that can be used for TSCD algorithm evaluation.
CCIL: Continuity-Based Data Augmentation for Corrective Imitation Learning
Liyiming Ke (University of Washington), Abhishek Gupta (University of Washington)
Autonomous DrivingRobotic IntelligenceReinforcement Learning from Human FeedbackReinforcement LearningAgentic AISequential
🎯 What it does: A data augmentation method is proposed that utilizes local continuity of environmental dynamics to generate corrective labels, enhancing the robustness of imitation learning.
CellPLM: Pre-training of Cell Language Model Beyond Single Cells
Hongzhi Wen (Michigan State University), Jiliang Tang (Michigan State University)
TransformerBiomedical Data
🎯 What it does: A cell language model called CellPLM based on Transformer is proposed, which can learn intercellular relationships through pre-training and generate high-quality cell embeddings.
Certified Adversarial Robustness for Rate Encoded Spiking Neural Networks
Bhaskar Mukhoty (Mohamed bin Zayed University of Artificial Intelligence), Bin Gu (Jilin University)
Adversarial AttackSpiking Neural NetworkImage
🎯 What it does: This paper studies the robustness of spike neural networks (SNN) using rate coding and provides a provable robustness proof under the l1-norm based on Bernoulli random smoothing.
Chain of Hindsight aligns Language Models with Feedback
Hao Liu (University of California Berkeley), Pieter Abbeel (University of California Berkeley)
GenerationReinforcement Learning from Human FeedbackTransformerLarge Language ModelSupervised Fine-TuningText
🎯 What it does: A novel method called 'Chain of Hindsight (CoH)' is proposed, which fine-tunes large language models by converting all human feedback into natural language sequences and using them as contextual input, enabling the model to learn and align with both positive and negative feedback.
Chain of Log-Concave Markov Chains
Saeed Saremi (Genentech), Francis Bach (Inria)
Gaussian SplattingStochastic Differential Equation
🎯 What it does: This paper proposes a non-Markov chain sampling framework that transforms sampling of unnormalized densities into a series of log-concave conditional density samples by using Gaussian kernel smoothing at a fixed noise scale and accumulating multiple noise measurements.
Chain of Thought Empowers Transformers to Solve Inherently Serial Problems
Zhiyuan Li (Toyota Technological Institute at Chicago and Stanford University), Tengyu Ma (Stanford University)
TransformerSequentialChain-of-Thought
🎯 What it does: This paper theoretically and experimentally demonstrates that introducing chain-of-thought (CoT) in decoder-only transformers can significantly enhance the model's expressive capability for inherently serial computation problems.
Chain-of-Experts: When LLMs Meet Complex Operations Research Problems
Ziyang Xiao (Zhejiang University), Gang Chen (Huawei Noah's Ark Lab)
OptimizationTransformerLarge Language ModelPrompt EngineeringMixture of ExpertsTextChain-of-Thought
🎯 What it does: The Chain-of-Experts (CoE) framework is proposed, utilizing multiple expert LLMs for collaborative reasoning to automatically complete the modeling and program generation of complex operations research (OR) problems.
Chain-of-Knowledge: Grounding Large Language Models via Dynamic Knowledge Adapting over Heterogeneous Sources
Xingxuan Li (Alibaba Group), Lidong Bing (Alibaba Group)
TransformerLarge Language ModelTextTabularBiomedical DataPhysics RelatedRetrieval-Augmented GenerationChain-of-Thought
🎯 What it does: This paper proposes the Chain-of-Knowledge (CoK) framework, which enhances the factual accuracy of large language models through a three-stage dynamic knowledge adaptation, utilizing multi-source (structured and unstructured) knowledge.
Chain-of-Table: Evolving Tables in the Reasoning Chain for Table Understanding
Zilong Wang (University of California), Tomas Pfister (Google Cloud AI Research)
TransformerLarge Language ModelPrompt EngineeringTabularBenchmarkChain-of-Thought
🎯 What it does: The CHAIN-OF-TABLE framework is proposed, which utilizes large language models to gradually perform table operations during the reasoning process, forming a table reasoning chain.
Chameleon: Increasing Label-Only Membership Leakage with Adaptive Poisoning
Harsh Chaudhari (Northeastern University), Jonathan Ullman (Northeastern University)
Safty and PrivacyAdversarial AttackConvolutional Neural NetworkImageTabular
🎯 What it does: This paper proposes a membership inference attack called Chameleon, which targets machine learning models that only return hard labels. The attack significantly improves the true positive rate at low FPR through adaptive poisoning and query strategies.
Channel Vision Transformers: An Image Is Worth 1 x 16 x 16 Words
Yujia Bao (Accenture), Theofanis Karaletsos (Chan Zuckerberg Initiative)
ClassificationRecognitionTransformerImage
🎯 What it does: Proposes ChannelViT, an improved Vision Transformer to handle multi-channel images, and enhances robustness using hierarchical channel sampling.
ChatEval: Towards Better LLM-based Evaluators through Multi-Agent Debate
Chi-Min Chan (Tsinghua University), Zhiyuan Liu (Tsinghua University)
TransformerLarge Language ModelPrompt EngineeringTextBenchmark
🎯 What it does: A multi-agent debate framework called ChatEval is proposed, which uses LLMs with different roles to collaboratively assess text quality in conversations.
CIFAR-10-Warehouse: Broad and More Realistic Testbeds in Model Generalization Analysis
Xiaoxiao Sun (Australian National University), Liang Zheng (Australian National University)
ClassificationDomain AdaptationDiffusion modelImageBenchmark
🎯 What it does: This paper proposes CIFAR-10-Warehouse, a dataset containing 180 categories that are the same as CIFAR-10 but sourced from different real or generated domains, aimed at evaluating the model's generalization ability in unseen environments.
Circuit Component Reuse Across Tasks in Transformer Language Models
Jack Merullo (Brown University), Ellie Pavlick (Brown University)
TransformerLarge Language ModelText
🎯 What it does: A study on circuit reuse in the GPT-2 model demonstrates that different tasks (IOI and Colored Objects) share approximately 78% of attention heads, and interventions improve the accuracy of the Colored Objects task.
CircuitNet 2.0: An Advanced Dataset for Promoting Machine Learning Innovations in Realistic Chip Design Environment
Xun Jiang (Peking University), Ru Huang (Peking University)
Graph Neural NetworkMultimodalityBenchmark
🎯 What it does: The CircuitNet 2.0 dataset is proposed, containing over 10,000 samples of CPUs, GPUs, and AI chips under a 14 nm FinFET process, and providing multi-modal features and multi-task labels.
Circumventing Concept Erasure Methods For Text-To-Image Generative Models
Minh Pham (New York University), Chinmay Hegde (New York University)
GenerationAdversarial AttackTransformerPrompt EngineeringDiffusion modelImageText
🎯 What it does: This study investigates seven methods of concept erasure in text-to-image models and designs a Concept Inversion (CI) attack, demonstrating that these methods cannot completely remove the target concept.
CivRealm: A Learning and Reasoning Odyssey in Civilization for Decision-Making Agents
Siyuan Qi (National Key Laboratory of General Artificial Intelligence, BIGAI), Song-Chun Zhu (National Key Laboratory of General Artificial Intelligence, BIGAI)
Recurrent Neural NetworkTransformerLarge Language ModelReinforcement Learning
🎯 What it does: In this paper, the authors propose CivRealm—an interactive environment based on the Civilization game, designed to simultaneously test the learning and reasoning abilities of agents, and provide two types of interfaces: tensor-based RL and LLM-based;
CLaM-TTS: Improving Neural Codec Language Model for Zero-Shot Text-to-Speech
Jaehyeon Kim (KRAFTON), Jaewoong Cho (KRAFTON)
GenerationData SynthesisTransformerLarge Language ModelAudio
🎯 What it does: CLaM-TTS is proposed, a zero-shot speech synthesis system that generates multi-dimensional code streams using probabilistic residual vector quantization (RVQ) and language models.
CLAP: Collaborative Adaptation for Patchwork Learning
Sen Cui (Tsinghua University), Fei Wang (Cornell University)
Federated LearningRepresentation LearningAuto EncoderMultimodalityBiomedical Data
🎯 What it does: A Patchwork Learning framework (CLAP) is designed to complete multimodal missing data imputation and representation learning under the condition that different clients have different combinations of modalities and do not share raw data.
Class Incremental Learning via Likelihood Ratio Based Task Prediction
Haowei Lin (Peking University), Bing Liu (University of Illinois at Chicago)
ClassificationRecognitionTransformerImage
🎯 What it does: A task recognition method based on likelihood ratios (TPL) is proposed, achieving class incremental learning without task numbering;
Class Probability Matching with Calibrated Networks for Label Shift Adaption
Hongwei Wen (University of Twente), Hanyuan Hang (University of Twente)
Domain AdaptationImage
🎯 What it does: A Class Probability Matching framework is proposed for matching on label probabilities, and based on this framework, the CPMC algorithm is designed to address the label shift domain adaptation problem.
Classification with Conceptual Safeguards
Hailey Joren (University of California San Diego), Berk Ustun (University of California San Diego)
ClassificationTabular
🎯 What it does: Introducing the 'Conceptual Safeguard' mechanism, which achieves safe selective classification in the conceptual bottleneck model by predicting concepts and propagating their uncertainties, thereby improving coverage while meeting the minimum accuracy requirement.
Cleanba: A Reproducible and Efficient Distributed Reinforcement Learning Platform
Shengyi Huang (Drexel University), Santiago Ontanon (Hugging Face)
Reinforcement LearningImage
🎯 What it does: A distributed reinforcement learning platform called Cleanba is proposed, addressing the reproducibility issues in traditional systems like IMPALA, and achieving reproducible, efficient, and scalable distributed PPO/IMPALA training on 57 Atari games.
CLEX: Continuous Length Extrapolation for Large Language Models
Guanzheng Chen (Sun Yat-sen University), Lidong Bing (Alibaba Group)
GenerationOptimizationTransformerLarge Language ModelSupervised Fine-TuningTextBenchmarkOrdinary Differential Equation
🎯 What it does: The CLEX method is proposed, which utilizes neural ODE to learn the continuous scale dynamics of position encoding (RoPE), allowing for the use of short sequences during training and supporting ultra-long contexts during testing, enabling the context window of LLMs to be expanded to more than four times the original training length without performance loss; this module is also designed as a lightweight component that can be directly inserted into RoPE.
Clifford Group Equivariant Simplicial Message Passing Networks
Cong Liu (University of Amsterdam), Patrick Forré (University of Amsterdam)
Graph Neural NetworkTime SeriesSequential
🎯 What it does: This paper proposes a Clifford algebra-based E(n)-equivariant Simplicial Message Passing Network (CSMPN) that performs geometric equivariant message passing on simplicial complexes.
ClimODE: Climate and Weather Forecasting with Physics-informed Neural ODEs
Yogesh Verma (Aalto University), Vikas Garg (Aalto University)
Time SeriesPhysics RelatedOrdinary Differential Equation
🎯 What it does: ClimODE is proposed, a climate and weather prediction model based on continuous-time neural ODEs, which directly implements the physical continuity equation, ensuring conservation of quantities and capable of producing uncertainty estimates.
CLIP the Bias: How Useful is Balancing Data in Multimodal Learning?
Ibrahim Alabdulmohsin (Google Deepmind), Xiaohua Zhai (Google Deepmind)
ClassificationRetrievalData-Centric LearningTransformerContrastive LearningImageMultimodality
🎯 What it does: The study uses data balancing techniques in CLIP pre-training to mitigate social biases, proposing an algorithm called Multi-Modal Moment Matching (M4) and conducting in-depth experiments on large-scale datasets to analyze the changes in model representation bias and association bias with training stages and data volume.
CLIP-MUSED: CLIP-Guided Multi-Subject Visual Neural Information Semantic Decoding
Qiongyi Zhou (Chinese Academy of Sciences), Huiguang He (Chinese Academy of Sciences)
TransformerMultimodalityBiomedical DataMagnetic Resonance Imaging
🎯 What it does: A CLIP-guided multi-subject visual neural information semantic decoding method (CLIP-MUSED) is proposed, which utilizes a Transformer to extract global brain signal features and guides token learning through the combination of low-level/high-level subject-specific tokens with Representational Similarity Analysis (RSA), achieving effective aggregation of multi-subject data and modeling of individual differences.
CLIPSelf: Vision Transformer Distills Itself for Open-Vocabulary Dense Prediction
Size Wu (Nanyang Technological University), Chen Change Loy (Nanyang Technological University)
Object DetectionSegmentationKnowledge DistillationRepresentation LearningTransformerContrastive LearningImage
🎯 What it does: This study investigates the dense feature representation of Vision Transformer (ViT) in CLIP and proposes CLIPSelf, which enhances the performance of this dense representation in open vocabulary dense prediction tasks through self-distillation.
Closing the Curious Case of Neural Text Degeneration
Matthew Finlayson (University of Southern California), Ashish Sabharwal (Allen Institute for Artificial Intelligence)
GenerationTransformerLarge Language ModelText
🎯 What it does: This paper explains why truncated sampling (such as nucleus and top-k) is effective in large language models and proposes a basis-aware truncation (BAT) sampling algorithm based on the softmax bottleneck.
Closing the Gap between TD Learning and Supervised Learning - A Generalisation Point of View.
Raj Ghugare (Mila), Benjamin Eysenbach (Princeton University)
TransformerSupervised Fine-TuningReinforcement LearningTime Series
🎯 What it does: This study investigates the 'stitching' property in RL, proving that the OCBC method based on supervised learning cannot achieve compositional generalization, and proposes time series data augmentation to address this deficiency; experiments are conducted on various state-based and image-based offline RL tasks.
CNN Kernels Can Be the Best Shapelets
Eric Qu (University of California), Dongsheng Li (Microsoft Research Asia)
ClassificationExplainability and InterpretabilityConvolutional Neural NetworkTime SeriesElectrocardiogram
🎯 What it does: Proposes ShapeConv - an interpretable CNN layer that uses convolution kernels as shapelets for end-to-end time series classification and clustering tasks.
CO2: Efficient Distributed Training with Full Communication-Computation Overlap
Weigao Sun (OpenNLPLab Shanghai AI Laboratory), Yiran Zhong (OpenNLPLab Shanghai AI Laboratory)
OptimizationComputational EfficiencyImageText
🎯 What it does: Proposes the CO2 method, achieving complete overlap of communication and computation in distributed training, significantly enhancing the scalability of low-bandwidth clusters.
CoBIT: A Contrastive Bi-directional Image-Text Generation Model
Haoxuan You (Columbia University), Jiahui Yu (Google Research)
GenerationRetrievalTransformerVision Language ModelContrastive LearningImageTextMultimodality
🎯 What it does: A unified contrastive bidirectional image-text generation model CoBIT is proposed, which can simultaneously train three pre-training objectives: image-text contrastive learning, image-to-text generation, and text-to-image generation within the same framework.
COCO-Periph: Bridging the Gap Between Human and Machine Perception in the Periphery
Anne Harrington (Massachusetts Institute of Technology), Ruth Rosenholtz (Massachusetts Institute of Technology)
Object DetectionConvolutional Neural NetworkTransformerImage
🎯 What it does: This paper constructs the COCO-Periph dataset, simulating human peripheral vision by applying a Uniform Texture Tiling Model (Uniform TTM) to MS-COCO images, and trains and evaluates various object detection models based on this; it also designs a machine psychophysical experiment to compare the performance of humans and models in peripheral object detection.
CODE REPRESENTATION LEARNING AT SCALE
Dejiao Zhang (Amazon), Bing Xiang (Amazon)
RetrievalRepresentation LearningTransformerContrastive LearningTextMultimodality
🎯 What it does: A bidirectional encoder model for source code, CODESAGE, has been constructed, and a two-stage pre-training scheme has been proposed: the first stage uses a combination of full masking (Full Mask) and identifier deobfuscation (DOBF) for masked language modeling; the second stage employs dual-modal contrastive learning (Text-Code pairs) and introduces hard positive and hard negative samples.
CodeChain: Towards Modular Code Generation Through Chain of Self-revisions with Representative Sub-modules
Hung Le (Salesforce Research), Shafiq Joty (Salesforce Research)
GenerationAI Code AssistantTransformerLarge Language ModelPrompt EngineeringTextBenchmarkChain-of-Thought
🎯 What it does: Proposes the CodeChain framework, which utilizes chain-of-thought (CoT) prompts to generate modular code and enhances code quality through a self-revision loop of representative submodules.
Coeditor: Leveraging Repo-level Diffs for Code Auto-editing
Jiayi Wei (Augment Computing), Isil Dillig (University of Texas at Austin)
AI Code AssistantTransformerSupervised Fine-TuningText
🎯 What it does: This paper studies the multi-round code auto-editing task based on code change history and proposes the Coeditor model for implementation.
COLEP: Certifiably Robust Learning-Reasoning Conformal Prediction via Probabilistic Circuits
Mintong Kang (University of Illinois Urbana-Champaign), Bo Li (University of Chicago)
Autonomous DrivingAdversarial AttackImage
🎯 What it does: This paper proposes the COLEP framework, which combines knowledge-driven logical reasoning (using probabilistic circuits) with split-based conformal prediction. It provides a global coverage proof for finite L₂ error and validates its robustness through experiments.
CoLiDE: Concomitant Linear DAG Estimation
Seyed Saman Saboksayr (University of Rochester), Mariano Tepper (Intel Labs)
Graph Neural NetworkGraphTabular
🎯 What it does: The CoLiDE framework is proposed, which simultaneously learns the linear DAG structure and the noise variance of each node through continuous optimization.
COLLIE: Systematic Construction of Constrained Text Generation Tasks
Shunyu Yao (Princeton University), Karthik R Narasimhan
GenerationTransformerLarge Language ModelPrompt EngineeringText
🎯 What it does: A grammar-based framework called COLLIE is proposed for systematically constructing multi-level and multi-type constraint text generation tasks.
Combinatorial Bandits for Maximum Value Reward Function under Value-Index Feedback
Yiliu Wang (Allen Institute), Milan Vojnovic (London School of Economics)
OptimizationReinforcement Learning from Human FeedbackReinforcement LearningTabular
🎯 What it does: Proposed a method to learn and maximize rewards in the k-MAX combinatorial multi-armed bandit problem using only feedback from the maximum value and the corresponding arm index;
Combining Axes Preconditioners through Kronecker Approximation for Deep Learning
Sai Surya Duvvuri (University of Texas at Austin), Inderjit S Dhillon
OptimizationGraph Neural NetworkTransformerTextGraph
🎯 What it does: This paper proposes CASPR, which utilizes Kronecker and Sum approximations to construct an axis preconditioner for matrix-type DNN parameters, thereby more accurately approximating the gradient second moment of the full matrix Adagrad.
Communication-Efficient Federated Non-Linear Bandit Optimization
Chuanhao Li (Yale University), Yu-Xiang Wang (University of California, Santa Barbara)
OptimizationFederated Learning
🎯 What it does: Proposes the Fed-GO-UCB algorithm to solve the federated team optimization problem, supporting general nonlinear objective functions, and proves sublinear cumulative regret and communication costs.
Communication-Efficient Gradient Descent-Accent Methods for Distributed Variational Inequalities: Unified Analysis and Local Updates
Siqi Zhang (Johns Hopkins University), Nicolas Loizou (Johns Hopkins University)
OptimizationFederated LearningReinforcement LearningTabular
🎯 What it does: This paper proposes a communication-efficient local training method for distributed variational inequality (VIP) problems and provides a unified convergence analysis.
CompA: Addressing the Gap in Compositional Reasoning in Audio-Language Models
Sreyan Ghosh (University of Maryland), Dinesh Manocha (University of Maryland)
ClassificationRetrievalLarge Language ModelContrastive LearningMultimodalityBenchmarkAudio
🎯 What it does: Two expert-annotated benchmarks, CompA-order and CompA-attribute, were proposed to evaluate the capability of audio-language models (ALM) in combinatorial reasoning. Based on this, the CompA-CLAP model was developed, enhancing the combinatorial reasoning ability of ALM through an improved contrastive learning method.
Complete and Efficient Graph Transformers for Crystal Material Property Prediction
Keqiang Yan (Texas A&M University), Shuiwang Ji (Texas A&M University)
Graph Neural NetworkTransformerGraphPhysics Related
🎯 What it does: The study investigates the prediction of crystal material properties and proposes a geometrically complete crystal graph representation along with the ComFormer model.
Complex priors and flexible inference in recurrent circuits with dendritic nonlinearities
Benjamin S. H. Lyo (New York University), Cristina Savin (New York University)
Spiking Neural NetworkDiffusion modelImage
🎯 What it does: A sampleable neural network utilizing dendritic branch neurons and global oscillations has been constructed, capable of implicitly representing complex priors and achieving flexible posterior inference in perceptual tasks.
Compose and Conquer: Diffusion-Based 3D Depth Aware Composable Image Synthesis
Jonghyun Lee (Korea University), Yonghyun Jeong (NAVER Cloud)
GenerationData SynthesisDepth EstimationDiffusion modelImage
🎯 What it does: A new conditional diffusion model called Compose and Conquer (CnC) is proposed, which can place multiple objects in 3D depth space and inject global semantics from multiple sources into local areas.
Composed Image Retrieval with Text Feedback via Multi-grained Uncertainty Regularization
Yiyang Chen (National University of Singapore), Tat-Seng Chua (National University of Singapore)
RetrievalConvolutional Neural NetworkContrastive LearningImageText
🎯 What it does: A multi-granularity uncertainty regularization method is proposed for combining fine-grained one-to-one matching with coarse-grained one-to-many matching in synthetic image retrieval.
Compositional Conservatism: A Transductive Approach in Offline Reinforcement Learning
Yeda Song (Seoul National University), Gunhee Kim (Seoul National University)
Reinforcement LearningTabularBenchmark
🎯 What it does: A conservative framework COCOA for offline reinforcement learning is proposed, which decomposes the state into anchor and delta, and utilizes an inverse dynamics model for anchor pathfinding, thereby enhancing the generalization ability of offline RL.
Compositional Generative Inverse Design
Tailin Wu (Westlake University), Jure Leskovec (Stanford University)
GenerationOptimizationDiffusion modelTime SeriesPhysics Related
🎯 What it does: This paper proposes a reverse design framework based on diffusion models, called CinDM, which utilizes an energy function for joint optimization of design variables and achieves extrapolation of time, object quantity, and geometric shapes by combining different sub-models.
Compositional Preference Models for Aligning LMs
Dongyoung Go (Naver Corporation), Marc Dymetman (Independent Researcher)
Recommendation SystemExplainability and InterpretabilityReinforcement Learning from Human FeedbackTransformerLarge Language ModelPrompt EngineeringText
🎯 What it does: Proposed and experimented with the 'Combination Preference Model (CPM)', which breaks down single preference evaluations into several interpretable features, then uses LM to extract feature scores and combines them into a final preference score through logistic regression;
Compressed Context Memory for Online Language Model Interaction
Jang-Hyun Kim (Seoul National University), Hyun Oh Song (Seoul National University)
CompressionRecommendation SystemTransformerLarge Language ModelText
🎯 What it does: A compressed context memory (CCM) framework is proposed, which supports dynamic context compression during online inference by inserting lightweight conditional LoRA for adaptive compression of keys/values in the Transformer.
Compressing Latent Space via Least Volume
Qiuyi Chen (University of Maryland), Mark Fuge (University of Maryland)
CompressionRepresentation LearningAuto EncoderImage
🎯 What it does: A regularization method called Least Volume is proposed, which can compress the latent space dimension of the autoencoder into a low-dimensional linear subspace without requiring prior information.
Compressing LLMs: The Truth is Rarely Pure and Never Simple
AJAY KUMAR JAISWAL, Yinfei Yang (Apple)
RetrievalCompressionTransformerLarge Language ModelTextBenchmark
🎯 What it does: This paper explores the true performance of compressed large language models in terms of knowledge retention, reasoning, retrieval, summarization, and instruction following by creating a new evaluation benchmark called LLM-KICK.
Concept Bottleneck Generative Models
Aya Abdelsalam Ismail (Genentech), Kyunghyun Cho (New York University)
GenerationExplainability and InterpretabilityDiffusion modelAuto EncoderGenerative Adversarial NetworkImage
🎯 What it does: A universal Concept Bottleneck (CB) layer is proposed, which can be inserted into VAE, GAN, and diffusion models, aligning the internal representations of generative models with human-interpretable concepts, thus achieving interpretable and adjustable generation.
Conditional Information Bottleneck Approach for Time Series Imputation
MinGyu Choi (Massachusetts Institute of Technology), Changhee Lee (Chung-Ang University)
Auto EncoderContrastive LearningImageTime SeriesSequentialBiomedical Data
🎯 What it does: A framework for time series missing value imputation based on Conditional Information Bottleneck (CIB) is proposed and implemented, addressing the shortcomings of traditional information bottleneck in capturing temporal dynamics.
Conditional Instrumental Variable Regression with Representation Learning for Causal Inference
Debo Cheng (University of South Australia), Thuc Duy Le (University of South Australia)
Representation LearningTabular
🎯 What it does: A nonlinear causal effect estimation method based on Conditional Instrumental Variables (CIV) called CBRL.CIV is proposed to eliminate observed and unobserved confounding bias and balance the distribution of observed covariates.
Conditional Variational Diffusion Models
Gabriel Della Maggiora, Artur Yakimovich (Helmholtz Zentrum Dresden Rossendorf)
RestorationGenerationSuper ResolutionDiffusion modelImageBiomedical Data
🎯 What it does: A conditional variational diffusion model (CVDM) is proposed, which can learn noise scheduling during training to achieve high-quality sampling for inverse problems.
Confidence-aware Reward Optimization for Fine-tuning Text-to-Image Models
Kyuyoung Kim (KAIST), Kimin Lee (Google DeepMind)
Image TranslationOptimizationReinforcement Learning from Human FeedbackTransformerReinforcement LearningVision Language ModelImageTextBenchmark
🎯 What it does: Confidence calibration of reward optimization for text-to-image models to reduce alignment and quality degradation caused by over-optimization.
Confidential-DPproof: Confidential Proof of Differentially Private Training
Ali Shahin Shamsabadi (Brave Software), Adrian Weller (University of Cambridge)
Safty and PrivacyComputational EfficiencyConvolutional Neural NetworkContrastive LearningImage
🎯 What it does: This paper proposes the Confidential-DPproof framework, which utilizes zero-knowledge proofs to generate a complete differential privacy (DP) proof during model training, providing verifiable credentials to external auditors regarding the ε and δ bounds achieved by DP-SGD training, without disclosing any training data or model information.
Conformal Inductive Graph Neural Networks
Soroush H. Zargarbashi (CISPA Helmholtz Center for Information Security), Aleksandar Bojchevski (University of Cologne)
ClassificationGraph Neural NetworkSupervised Fine-TuningGraph
🎯 What it does: A new framework combining conformal prediction and graph neural networks (GNN) for the task of inducing node/edge classification is proposed: NodeEx-CP and EdgeEx-CP, which can dynamically recalculate embeddings and compliance scores as the graph evolves, restoring coverage guarantees.
Conformal Language Modeling
Victor Quach (Massachusetts Institute of Technology), Regina Barzilay (Massachusetts Institute of Technology)
GenerationExplainability and InterpretabilityTransformerLarge Language ModelTextMagnetic Resonance Imaging
🎯 What it does: Provides a reliable prediction set for generative language models without distribution and with limited sample size, and can further select high-confidence components at the clause or sentence level.
Conformal Prediction via Regression-as-Classification
Etash Kumar Guha, Eugene Ndiaye (Apple)
ClassificationOptimizationTabular
🎯 What it does: This paper proposes a conformal prediction method that transforms regression problems into classification problems (R2CCP), by discretizing the output space and using a custom loss to learn the probability distribution, thereby obtaining prediction intervals with reliable coverage.
Conformal Risk Control
Anastasios Nikolas Angelopoulos, Tal Schuster (Stanford University)
ClassificationSegmentationImage
🎯 What it does: A general conformal-style risk control method is proposed, which can provide expected value control guarantees for any monotonic loss function.
Confronting Reward Model Overoptimization with Constrained RLHF
Ted Moskovitz (University College London), Stephen Marcus McAleer
OptimizationReinforcement Learning from Human FeedbackTransformerReinforcement LearningText
🎯 What it does: The study examines the performance of reward models under excessive optimization in composite rewards and proposes a RLHF method based on constrained reinforcement learning to avoid over-optimization.
ConjNorm: Tractable Density Estimation for Out-of-Distribution Detection
Bo Peng (University of Technology Sydney), Zhen Fang (University of Technology Sydney)
Anomaly DetectionConvolutional Neural NetworkImage
🎯 What it does: A theoretical framework based on Bregman divergence is proposed, transforming OOV detection into the search for the optimal density function on exponential family distributions, and the CONJNORM method is provided; at the same time, importance sampling is introduced to achieve an unbiased estimate of the normalization constant, completing a comprehensive post-hoc OOV detection scheme.
Connect, Collapse, Corrupt: Learning Cross-Modal Tasks with Uni-Modal Data
Yuhui Zhang (Stanford University), Serena Yeung
GenerationData SynthesisRetrievalTransformerLarge Language ModelContrastive LearningImageVideoTextMultimodalityAudio
🎯 What it does: Through theoretical analysis of the geometric space of multimodal contrastive learning, a three-step method called C3 (Connect‑Collapse‑Corrupt) is proposed, enabling the training of cross-modal tasks using single-modal data, achieving state-of-the-art performance in zero-shot image captioning, audio/video captioning, and text-to-image generation tasks.
Connecting Large Language Models with Evolutionary Algorithms Yields Powerful Prompt Optimizers
Qingyan Guo (Tsinghua University), Yujiu Yang (Tsinghua University)
OptimizationTransformerLarge Language ModelPrompt EngineeringText
🎯 What it does: By combining evolutionary algorithms with large language models (LLM), this paper automates the optimization of discrete prompts, generating readable and more effective prompts.
ConR: Contrastive Regularizer for Deep Imbalanced Regression
Mahsa Keramati (Simon Fraser University), R. David Evans (Borealis AI)
Depth EstimationContrastive LearningImage
🎯 What it does: A contrastive regularizer for deep imbalance regression, ConR, is proposed, which utilizes local and global similarities in the continuous label space to regularize the feature space and prevent feature collapse of minority samples.
Consciousness-Inspired Spatio-Temporal Abstractions for Better Generalization in Reinforcement Learning
Harry Zhao, Yoshua Bengio (Mila)
Reinforcement Learning
🎯 What it does: Proposes the Skipper framework, which uses spatiotemporal abstraction to automatically generate proxy problems, achieving zero-shot generalization reinforcement learning through checkpoints in decision-making planning.
Conserve-Update-Revise to Cure Generalization and Robustness Trade-off in Adversarial Training
Shruthi Gowda (Eindhoven University of Technology), Elahe Arani
ClassificationAdversarial AttackConvolutional Neural NetworkGenerative Adversarial NetworkImage
🎯 What it does: The CURE framework is proposed, achieving a better balance between adversarial robustness and standard generalization through dynamic retention, updating, and revision of weights in adversarial training.
Consistency Models as a Rich and Efficient Policy Class for Reinforcement Learning
Zihan Ding (Princeton University), Chi Jin (Princeton University)
Reinforcement LearningDiffusion modelTabularBenchmarkOrdinary Differential Equation
🎯 What it does: The researchers use a consistency model as a policy representation in reinforcement learning, proposing two algorithms: Consistency-BC (Behavior Cloning) and Consistency-AC (Actor-Critic), and evaluate them in offline, offline-to-online, and online RL scenarios.
Consistency Training with Learnable Data Augmentation for Graph Anomaly Detection with Limited Supervision
Nan Chen (National University of Singapore), Jia Chen (GrabTaxi Holdings Pte. Ltd.)
Anomaly DetectionGraph Neural NetworkContrastive LearningGraph
🎯 What it does: In response to Graph Anomaly Detection (GAD) in scenarios with limited supervision, the CONSISGAD model is proposed, which utilizes consistency training and learnable data augmentation to fully leverage unlabeled data, and enhances detection performance through a GNN structure driven by homophily distribution.
Consistency Trajectory Models: Learning Probability Flow ODE Trajectory of Diffusion
Dongjun Kim (Sony AI), Stefano Ermon (Stanford University)
GenerationData SynthesisDiffusion modelScore-based ModelGenerative Adversarial NetworkImageOrdinary Differential Equation
🎯 What it does: This paper proposes the Consistency Trajectory Model (CTM), a generative framework that simultaneously considers score-based models and distillation models, capable of achieving probability flow ODE jumps at arbitrary time intervals in a single forward pass, and supports controllable sampling.
Consistency-guided Prompt Learning for Vision-Language Models
Shuvendu Roy (Queen's University), Ali Etemad (Queen's University)
ClassificationDomain AdaptationTransformerLarge Language ModelPrompt EngineeringVision Language ModelImageTextMultimodality
🎯 What it does: In the few-shot fine-tuning of visual-language models (such as CLIP), a method called Consistency-Guided Prompt Learning (CoPrompt) is proposed, which enhances the model's generalization ability in downstream tasks and zero-shot tasks by utilizing consistency constraints, input perturbations, and a combination of prompts and adapters.
Consistent algorithms for multi-label classification with macro-at-$k$ metrics
Erik Schultheis (Aalto University), Krzysztof Dembczynski (Yahoo Research)
ClassificationOptimizationBenchmark
🎯 What it does: A consistency learning framework for macro-average (macro-atk) metrics that predicts only k labels for each instance in multi-label classification is proposed, and the analytical form of the optimal classifier is provided.
Consistent Multi-Class Classification from Multiple Unlabeled Datasets
Zixi Wei (Chongqing University), Lei Feng (Nanyang Technological University)
ClassificationConvolutional Neural NetworkImage
🎯 What it does: Proposes a method for learning multi-class classifiers given multiple sets of unlabeled data and their class prior probabilities.
Consistent Video-to-Video Transfer Using Synthetic Dataset
Jiaxin Cheng (Amazon Web Services), Tong He (Amazon Web Services)
GenerationData SynthesisPrompt EngineeringDiffusion modelOptical FlowVideoText
🎯 What it does: A text-driven video-to-video editing method based on synthetic datasets (InsV2V) is proposed, achieving single model coverage for all videos, no per-video fine-tuning, and editing can be completed with just editing prompts.
Consistent4D: Consistent 360° Dynamic Object Generation from Monocular Video
Yanqin Jiang (Chinese Academy of Sciences), Yao Yao (Nanjing University)
GenerationData SynthesisDiffusion modelNeural Radiance FieldVideo
🎯 What it does: A Consistent4D framework is proposed, which generates 360° 4D dynamic objects from single-view uncalibrated videos through Cascade DyNeRF and interpolation consistency loss.
Constrained Bi-Level Optimization: Proximal Lagrangian Value Function Approach and Hessian-free Algorithm
Wei Yao (Southern University of Science and Technology), Jin Zhang (Southern University of Science and Technology)
OptimizationFederated LearningTabular
🎯 What it does: A single-layer smoothing method based on the proximal Lagrangian value function is proposed, and on this basis, a Hessian-free, single-loop gradient optimization algorithm is designed to solve bi-level optimization problems with upper and lower layer coupling constraints.
Constrained Decoding for Cross-lingual Label Projection
Duong Minh Le (Georgia Institute of Technology), Wei Xu (Georgia Institute of Technology)
ClassificationRecognitionTransformerSupervised Fine-TuningText
🎯 What it does: A cross-language label projection method based on constrained decoding (CODEC) is proposed for automatically translating and projecting word/phrase-level labels in low-resource languages.
Constraint-Free Structure Learning with Smooth Acyclic Orientations
Riccardo Massidda (Università di Pisa), Davide Bacciu (Università di Pisa)
OptimizationGraph Neural NetworkGraph
🎯 What it does: A constraint-free continuous optimization framework COSMO is designed, utilizing smooth priority vectors and temperature annealing to learn DAGs.
Constructing Adversarial Examples for Vertical Federated Learning: Optimal Client Corruption through Multi-Armed Bandit
Duanyi YAO, Jin Liu (Hong Kong University of Science and Technology)
Federated LearningAdversarial AttackImageTabular
🎯 What it does: An adaptive adversarial attack for vertical federated learning is proposed, which can dynamically select compromised clients and generate adversarial samples during the attack process to maximize the attack success rate.
Context is Environment
Sharut Gupta (Meta AI, Massachusetts Institute of Technology), Kartik Ahuja (Meta AI)
Domain AdaptationTransformerLarge Language ModelImage
🎯 What it does: A new domain generalization method is proposed - In-Context Risk Minimization (ICRM), which treats the environment as context and dynamically focuses on minimizing the risk of the test environment using unlabeled contextual information during inference.
Context-Aware Meta-Learning
Christopher Fifty (Stanford University), Sebastian Thrun (Google DeepMind)
Meta LearningTransformerImage
🎯 What it does: Developed Context-Aware Meta Learning (CAML), which can learn new visual concepts without fine-tuning during inference.
ContextRef: Evaluating Referenceless Metrics for Image Description Generation
Elisa Kreiss (University of California Los Angeles), Nick Haber (Stanford University)
GenerationSupervised Fine-TuningVision Language ModelImageTextBenchmark
🎯 What it does: Proposed the ContextRef benchmark, which evaluates the quality of reference-free image descriptions using human assessment and data augmentation.
Contextual Bandits with Online Neural Regression
Rohan Deb (University of Illinois), Arindam Banerjee (University of Illinois)
Recommendation SystemOptimizationNeural Architecture SearchReinforcement LearningTabular
🎯 What it does: This paper proposes a technique for online regression using neural networks, and through the reduction of Foster–Rakhlin/Foster–Krishnamurthy, transforms the regret of online regression into the regret of contextual bandits. By adding small random perturbations to the network output, it ensures that the loss satisfies the conditions of almost-convexity, QG, and a unique minimum point, thereby achieving O(log T) online regression regret for squared loss and KL loss; subsequently, it obtains O(√KT) regret for NeuSquareCB and O(√(L*K)+K) regret for NeuFastCB.
Continual Learning in the Presence of Spurious Correlations: Analyses and a Simple Baseline
Donggyu Lee (Sungkyunkwan University), Taesup Moon (Seoul National University)
ClassificationData-Centric LearningImage
🎯 What it does: This study investigates the issue of bias transfer caused by dataset bias in continual learning, systematically evaluating its impact on forward and backward transfer, and proposes a general baseline method BGS based on group-class balanced greedy sampling.
Continual Learning on a Diet: Learning from Sparsely Labeled Streams Under Constrained Computation
Wenxuan Zhang (King Abdullah University of Science and Technology), Mohamed Elhoseiny (King Abdullah University of Science and Technology)
Computational EfficiencyTransformerAuto EncoderImage
🎯 What it does: This study investigates continuous learning scenarios with limited computational budgets and sparse labels, and proposes the DietCL baseline algorithm for efficiently utilizing both labeled and unlabeled data.
Continual Momentum Filtering on Parameter Space for Online Test-time Adaptation
Jae-Hong Lee (Hanyang University), Joon-Hyuk Chang (Hanyang University)
Domain AdaptationOptimizationImageStochastic Differential EquationAudio
🎯 What it does: A continuous momentum filtering (CMF) framework is proposed for online testing time adaptation (OTTA), which alternately updates the model between SGD optimization and parameter denoising using a Kalman filter.
Continuous Field Reconstruction from Sparse Observations with Implicit Neural Networks
Xihaier Luo (Brookhaven National Laboratory), Shinjae Yoo (Brookhaven National Laboratory)
Auto EncoderTime SeriesPhysics Related
🎯 What it does: A continuous field reconstruction model MMGN based on implicit neural representation (INR) is proposed, which can recover global space-time physical fields from sparse and time-varying observational data.
Continuous Invariance Learning
LIN Yong, Hao Wang (Rutgers University)
Domain AdaptationAdversarial AttackTabularTime Series
🎯 What it does: A new continuous domain invariance learning framework (CIL) is proposed to address the issue of traditional methods like IRM failing easily in continuous domains.
Continuous-Multiple Image Outpainting in One-Step via Positional Query and A Diffusion-based Approach
Shaofeng Zhang (Shanghai Jiao Tong University), Junchi Yan (Shanghai Jiao Tong University)
GenerationData SynthesisDiffusion modelImage
🎯 What it does: This paper proposes a new image extrapolation method called PQDiff, which can generate image content at any multiple in a single step, addressing two technical challenges that have not been solved in the existing literature: 1) extrapolation at arbitrary continuous multiples, and 2) completing extrapolation in one step.
Contrastive Difference Predictive Coding
Chongyi Zheng (Carnegie Mellon University), Benjamin Eysenbach (Princeton University)
Robotic IntelligenceReinforcement LearningContrastive LearningImageSequential
🎯 What it does: A time-difference-based InfoNCE (TD-InfoNCE) learning framework is proposed and implemented to estimate the future state occupancy distribution of the target policy in offline or data-limited scenarios, thereby driving goal-conditioned reinforcement learning (Goal-Conditioned RL) algorithms.
Contrastive Learning is Spectral Clustering on Similarity Graph
Zhiquan Tan (Tsinghua University), Yang Yuan (Tsinghua University)
ClassificationRepresentation LearningContrastive LearningImageMultimodality
🎯 What it does: It is proven that SimCLR (InfoNCE) is equivalent to performing spectral clustering on the similarity graph generated by data augmentation, and this perspective is extended to CLIP in multimodal learning; a Kernel-InfoNCE loss is proposed based on the maximum entropy principle, using a mixed form of exponential kernel instead of Gaussian kernel, with experiments validating its performance superior to the original SimCLR on CIFAR-10/100 and TinyImageNet.