Is this you? Claim this profile to correct it, add a bio and choose the work people see first.
Claim this profileWorks182 from public data
- Learning deep structured semantic models for web search using clickthrough data2,323
A series of new latent semantic models with a deep structure that project queries and documents into a common low-dimensional space where the relevance of a document given a query is readily computed as the distance between them are developed.
- Oscar: Object-Semantics Aligned Pre-training for Vision-Language Tasks2,301
This paper proposes a new learning method Oscar (Object-Semantics Aligned Pre-training), which uses object tags detected in images as anchor points to significantly ease the learning of alignments.
- MS-Celeb-1M: A Dataset and Benchmark for Large-Scale Face Recognition2,241
A benchmark task to recognize one million celebrities from their face images, by using all the possibly collected face images of this individual on the web as training data, which could lead to one of the largest classification problems in computer vision.
- LLaVA-Med: Training a Large Language-and-Vision Assistant for Biomedicine in One Day2,195
This paper proposes a cost-efficient approach for training a vision-language conversational assistant that can answer open-ended research questions of biomedical images, and releases instruction-following data and the LLaVA-Med model, which exhibits excellent multimodal conversational capability.
- Stacked Attention Networks for Image Question Answering2,088
A multiple-layer SAN is developed in which an image is queried multiple times to infer the answer progressively, and the progress that the SAN locates the relevant visual clues that lead to the answer of the question layer-by-layer.
- Grounded Language-Image Pre-training1,831
A grounded language-image pretraining model for learning object-level, language-aware, and semantic-rich visual representations that unifies object detection and phrase grounding for pre-training and can leverage massive image-text pairs by generating grounding boxes in a self-training fashion.
- From captions to visual concepts and back1,363
This paper uses multiple instance learning to train visual detectors for words that commonly occur in captions, including many different parts of speech such as nouns, verbs, and adjectives, and develops a maximum-entropy language model.
- 1,137
- A Persona-Based Neural Conversation Model1,122
This work presents persona-based models for handling the issue of speaker consistency in neural response generation that yield qualitative performance improvements in both perplexity and BLEU scores over baseline sequence-to-sequence models.
- Unified Vision-Language Pre-Training for Image Captioning and VQA1,084
VLP is the first reported model that achieves state-of-the-art results on both vision-language generation and understanding tasks, as disparate as image captioning and visual question answering, across three challenging benchmark datasets: COCO Captions, Flickr30k Captions and VQA 2.0.
- GLIGEN: Open-Set Grounded Text-to-Image Generation1,027
A novel approach that builds upon and extends the functionality of existing pre-trained text-to-image diffusion models by enabling them to also be conditioned on grounding inputs, and achieves open-world grounded text2img generation with caption and bounding box condition inputs.
- RegionCLIP: Region-based Language-Image Pretraining1,000
A new method is proposed that signifi-cantly extends CLIP to learn region-level visual representations, thus enabling fine-grained alignment between image regions and textual concepts, and the learned region representations support zero-shot inference for object detection, showing promising results on both COCO and LVIS datasets.
- Sora: A Review on Background, Technology, Limitations, and Opportunities of Large Vision Models734
This paper presents a comprehensive review of the model's background, related technologies, applications, remaining challenges, and future directions of text-to-video AI models.
- Reinforced Cross-Modal Matching and Self-Supervised Imitation Learning for Vision-Language Navigation660
A novel Reinforced Cross-Modal Matching (RCM) approach that enforces cross-modal grounding both locally and globally via reinforcement learning (RL), and a Self-Supervised Imitation Learning (SIL) method to explore unseen environments by imitating its own past, good decisions is introduced.
- 600
- Domain Adaptation via Pseudo In-Domain Data Selection584
The results show that more training data is not always better, and that best results are attained via proper domain-relevant data selection, as well as combining in- and general-domain systems during decoding.
- Adapting boosting for information retrieval measures575
This work presents a new ranking algorithm that combines the strengths of two previous methods: boosted tree classification, and LambdaRank, and shows how to find the optimal linear combination for any two rankers, and uses this method to solve the line search problem exactly during boosting.
- Semantic Compositional Networks for Visual Captioning485
Experimental results show that the proposed method significantly outperforms prior state-of-the-art approaches, across multiple evaluation metrics.
- Selecting good expansion terms for pseudo-relevance feedback480
This study re-examines the assumption that most frequent terms in the pseudo-feedback documents are useful for the retrieval and proposes to integrate a term classification process to predict the usefulness of expansion terms.
- 419
- Multi-Scale Vision Longformer: A New Vision Transformer for High-Resolution Image Encoding405
A comprehensive empirical study shows that the new ViT significantly outperforms several strong baselines, including the existing ViT models and their ResNet counterparts, and the Pyramid Vision Transformer from a concurrent work, on a range of vision tasks, including image classification, object detection, and segmentation.
- Generalized Decoding for Pixel, Image, and Language404
X-Decoder is the first work that provides a unified way to support all types of image segmentation and a variety of vision-language (VL) tasks with strong transferability to a wide range of downstream tasks in both zero-shot and finetuning settings.
- Towards Learning a Generic Agent for Vision-and-Language Navigation via Pre-Training388
This paper presents the first pre-training and fine-tuning paradigm for vision-and-language navigation (VLN) tasks, which leads to significant improvement over existing methods, achieving a new state of the art.
- TraceVLA: Visual Trace Prompting Enhances Spatial-Temporal Awareness for Generalist Robotic Policies353
A compact VLA model based on 4B Phi-3-Vision, pretrained on the Open-X-Embodiment and finetuned on the authors' dataset, rivals the 7B OpenVLA baseline while significantly improving inference efficiency.
- 332
- Object-Driven Text-To-Image Synthesis via Adversarial Training326
A thorough comparison between the classic grid attention and the new object-driven attention is provided through analyzing their mechanisms and visualizing their attention layers, showing insights of how the proposed model generates complex scenes in high quality.
- Tensor Programs V: Tuning Large Neural Networks via Zero-Shot Hyperparameter Transfer322
This work shows that, in the recently discovered Maximal Update Parametrization (muP), many optimal HPs remain stable even as model size changes, which leads to a new HP tuning paradigm, muTransfer: parametrize the target model in muP, tune the HP indirectly on a smaller model, and zero-shot transfer them to the full-sized model, i.e., without directly tuning the latter at all.
- StoryGAN: A Sequential Conditional GAN for Story Visualization312
A new story-to-image-sequence generation model, StoryGAN, based on the sequential conditional GAN framework is proposed, which outperformed state-of-the-art models in image quality, contextual consistency metrics, and human evaluation.
- 304
- Unified Contrastive Learning in Image-Text-Label Space301
This work intro-duce a new formulation of Unified Con-trastive Learning (UniCL) with a single learning objective to seamlessly prompt the synergy of two data types, universally for image recognition in zero-shot, linear-probing, fully finetuning and transfer learning sce-narios.
- Agent AI: Surveying the Horizons of Multimodal Interaction288
It is argued that by developing agentic AI systems in grounded environments, one can also mitigate the hallucinations of large foundation models and their tendency to generate environmentally incorrect outputs.
- Set-of-Mark Prompting Unleashes Extraordinary Visual Grounding in GPT-4V272
The experiments show that GPT-4V with SoM in zero-shot setting outperforms the state-of-the-art fully-finetuned referring expression comprehension and segmentation model on RefCOCOg, and the effectiveness of SoM on a wide range of fine-grained vision and multimodal tasks is validated.
- Semantic-SAM: Segment and Recognize Anything at Any Granularity272
Semantic-SAM, a universal image segmentation model to enable segment and recognize anything at any desired granularity, is introduced, representing the first attempt to jointly train a model on SA-1B, generic, and part segmentation datasets.
- LLaVA-Plus: Learning to Use Tools for Creating Multimodal Agents266
LLaVA-Plus is distinct in that the image query is directly grounded and actively engaged throughout the entire human-AI interaction sessions, significantly improving tool use performance and enabling new scenarios.
- 266
- KAT: A Knowledge Augmented Transformer for Vision-and-Language232
This work proposes a - Knowledge Augmented Transformer (KAT) - which achieves a strong state-of-the-art result on the open-domain multimodal task of OK-VQA and integrates implicit and explicit knowledge in an encoder-decoder architecture, while still jointly reasoning over both knowledge sources during answer generation.
- ELEVATER: A Benchmark and Toolkit for Evaluating Language-Augmented Visual Models211
This work builds ELEVATER (Evaluation of Language-augmented Visual Task-level Transfer), the first benchmark and toolkit for evaluating (pre-trained) language-AUgmented visual models.
- Tactical Rewind: Self-Correction via Backtracking in Vision-And-Language Navigation188
The Frontier Aware Search with backTracking (FAST) Navigator is presented, a general framework for action decoding, that achieves state-of-the-art results on the 2018 Room-to-Room (R2R) Vision-and-Language navigation challenge.
- BBQ-Networks: Efficient Exploration in Deep Reinforcement Learning for Task-Oriented Dialogue Systems174
A new algorithm is presented that significantly improves the efficiency of exploration for deep Q-learning agents in dialogue systems, and shows that spiking the replay buffer with experiences from just a few successful episodes can make Q- learning feasible when it might otherwise fail.
- GPT-4V in Wonderland: Large Multimodal Models for Zero-Shot Smartphone GUI Navigation166
The findings demonstrate that large multimodal models, specifically GPT-4V, excel in zero-shot GUI navigation through its advanced screen interpretation, action reasoning, and precise action localization capabilities.
- TACo: Token-aware Cascade Contrastive Learning for Video-Text Alignment165
A new algorithm called Token-Aware Cascade contrastive learning (TACo) is presented that improves Contrastive learning using two novel techniques, the token-aware contrastive loss which is computed by taking into account the syntactic classes of words.
- 156
- WebQA: Multihop and Multimodal QA156
This work introduces WEBQA, a challenging new benchmark that proves difficult for large-scale state-of-the-art models which lack language groundable visual representations for novel objects and the ability to reason, yet trivial for humans.
- Language-Based Image Editing with Recurrent Attentive Models151
A generic modeling framework for two subtasks of LBIE: language-based image segmentation and image colorization using recurrent attentive models to fuse image and language features is proposed.
- CollabLLM: From Passive Responders to Active Collaborators145
CollabLLM is introduced, a novel and general training framework that enhances multiturn human-LLM collaboration and estimates the long-term contribution of responses using Multiturn-aware Rewards by reinforcement fine-tuning these rewards.
- A Data Augmentation Approach for Sign-Language-To-Text Translation In-The-Wild135
The goal is to build a state-of-the-art system for the automatic interpretation of sign language in unpredictable video framing conditions and the set of hypotheses to build such a system are described and the list of experiments that will lead to their verification are described.
- 131
- 128
- 126
- Learning Continuous Phrase Representations for Translation Modeling123
This paper improves the performance of a state-of-the-art phrase-based statistical machine translation system trained on WMT 2012 French-English data by up to 1.3 BLEU points.
- Interactive Code Generation via Test-Driven User-Intent Formalization114
The workflow of interactive test-driven code generation is proposed, which leverages lightweight user feedback to formalize the user intent using generated tests that can be useful for debugging, and produce an improved set of code suggestions by pruning and ranking candidate code suggestions.
- Sequential Attention GAN for Interactive Image Editing114
A novel Sequential Attention Generative Adversarial Network (SeqAttnGAN) is proposed, which applies a neural state tracker to encode the previous image and the textual description in each turn of the sequence, and uses a GAN framework to generate a modified version of the image that is consistent with the preceding images and coherent with the description.
- Sequential Attention GAN for Interactive Image Editing via Dialogue.114
A novel Sequential Attention Generative Adversarial Network (SeqAttnGAN) is proposed, which applies a neural state tracker to encode the previous image and the textual description in each turn of the sequence, and uses a GAN framework to generate a modified version of the image that is consistent with the preceding images and coherent with the description.
- K-LITE: Learning Transferable Visual Models with External Knowledge112
This work proposes K-LITE, a simple strategy to leverage external knowledge for building transferable visual systems that enriches entities in text with WordNet and Wiktionary knowledge, leading to an efficient and scalable approach to learning image representations that uses knowledge about the visual concepts.
- TemporalBench: Benchmarking Fine-grained Temporal Understanding for Multimodal Video Models97
A new benchmark dedicated to evaluating fine-grained temporal understanding in videos, TemporalBench, which shows that state-of-the-art models like GPT-4o achieve only 38.5% question answering accuracy on TemporalBench, demonstrating a significant gap between humans and AI in temporal understanding.
- 96
- Learning Customized Visual Models with Retrieval-Augmented Knowledge93
This work retrieves the most relevant image-text pairs from the web-scale database as external knowledge and proposes React,REtrieval-Augmented CusTomization, a framework to acquire the relevant web knowledge to build customized visual models for target domains.
- 93
- Person resolution in person search results93
A user study shows that WebHawk complements most existing search engines and successfully improves users' experience of person search on the Web.
- A clinically accessible small multimodal radiology model and evaluation metric for chest X-ray findings91
LLaVA-Rad is a scalable, privacy-preserving and cost-effective step towards clinically adaptable multimodal AI for radiology, and CheXprompt, a radiologist-aligned factuality metric, to enable scalable, privacy-preserving image analysis.
- TIGEr: Text-to-Image Grounding for Image Caption Evaluation86
The empirical tests show that TIGEr has a higher consistency with human judgments than alternative existing metrics, and comprehensively assess the metric’s effectiveness in caption evaluation by measuring the correlation between human judgments and metric scores.
- 82
- 77
- What Makes A Good Story? Designing Composite Rewards for Visual Storytelling76
This paper proposes three assessment criteria: relevance, coherence and expressiveness, which are observed through empirical analysis could constitute a “high-quality” story to the human eye and proposes a reinforcement learning framework, ReCo-RL, with reward functions designed to capture the essence of these quality criteria.
- Visual in-Context Prompting75
This paper builds on top of an encoder-decoder architecture, and develops a versatile prompt encoder to support a variety of prompts like strokes, boxes, and points, and enhances it to take an arbitrary number of reference image segments as the context.
- 74
- 70
- VIVO: Visual Vocabulary Pre-Training for Novel Object Captioning64
The results show that the VIsual VOcabulary pre-training model can not only generate fluent image captions that describe novel objects, but also identify the locations of these objects.
- Fault-Aware Neural Code Rankers61
The proposed CodeRanker is a neural ranker that can predict the correctness of a sampled program without executing it, and can significantly increase the pass@1 accuracy of various code generation models (including Codex, GPT-Neo, G PT-J) on APPS, HumanEval and MBPP datasets.
- A Survey on Post-training of Large Language Models59
This paper presents the first comprehensive survey of PoLMs, systematically tracing their evolution across five core paradigms: Fine-tuning, which enhances task-specific accuracy; Alignment, which ensures ethical coherence and alignment with human preferences; Reasoning, which advances multi-step inference despite challenges in reward design; Efficiency, which optimizes resource utilization amidst increasing complexity; Integration and Adaptation, which extend capabilities across diverse modalities while addressing coherence issues.
- K-VIL: Keypoints-Based Visual Imitation Learning59
The keypoint-based visual imitation learning (K-VIL) approach is proposed, capable of extracting sparse, object-centric, and embodiment-independent task representations from a small number of human demonstration videos and of incrementally updating them when new demonstrations are available.
- VIVO: Surpassing Human Performance in Novel Object Captioning with Visual Vocabulary Pre-Training.59
The results show that the VIsual VOcabulary pre-training model can not only generate fluent image captions that describe novel objects, but also identify the locations of these objects.
- LLaVA-Interactive: An All-in-One Demo for Image Chat, Segmentation, Generation and Editing57
The development of LLaVA-Interactive is extremely cost-efficient as the system combines three multimodal skills of pre-built AI models without additional model training: visual chat of L LaVA, image segmentation from SEEM, as well as image generation and editing from GLIGEN.
- ValueNet: A New Dataset for Human Value Driven Dialogue System57
This work presents a new large-scale human value dataset called ValueNet, which contains human attitudes on 21,374 text scenarios and is the first one trying to incorporate a value model into emotionally intelligent dialogue systems.
- Is Your Goal-Oriented Dialog Model Performing Really Well? Empirical Analysis of System-wise Evaluation54
This paper performs a system-wise evaluation and presents an empirical analysis on different types of dialog systems which are composed of different modules in different settings to show that a pipeline dialog system trained using fine-grained supervision signals at different component levels often obtains better performance.
- Adaptive Chinese word segmentation53
A statistical framework where domain-specific words are identified in a unified approach to word segmentation based on linear models is presented and a transformation-based learning method used to adapt the system to differentword segmentation standards is described.
- Learning Phrase-Based Spelling Error Models from Clickthrough Data52
First, large amounts of query-correction pairs are derived by analyzing users' query reformulation behavior encoded in the clickthrough data, and a phrase-based error model is trained and integrated into a query speller system.
- 47
- Bi-KVIL: Keypoints-based Visual Imitation Learning of Bimanual Manipulation Tasks46
The proposed Bi-KVIL jointly extracts so-called Hybrid Master-Slave Relationships (HMSR) among objects and hands, bimanual coordination strategies, and sub-symbolic task representations, thus generalizing well to categorical objects in novel scenes.
- An Empirical Study of Scaling Instruct-Tuned Large Multimodal Models44
An empirical study of scaling LLaVA up to 33B and 65B/70B and performance of LoRA/QLoRA tuning of LMM are comparable to the performance of full-model fine-tuning, finding that scaling LMM consistently enhances model performance and improves language capabilities.
- 42
- Learning Lexicon Models from Search Logs for Query Expansion42
Evaluations on a real world data set show that the lexicon models, integrated into a ranker-based QE system, not only significantly improve the document retrieval performance but also outperform two state-of-the-art log- based QE methods.
- FlowRL: Matching Reward Distributions for LLM Reasoning41
This work transforms scalar rewards into a normalized target distribution using a learnable partition function, and then minimize the reverse KL divergence between the policy and the target distribution to promote diverse exploration and generalizable reasoning trajectories.
- Decoder Integration and Expected BLEU Training for Recurrent Neural Network Language Models41
This work shows how a recurrent neural network language model can be optimized towards an expected BLEU loss instead of the usual cross-entropy criterion, and tackles the issue of directly integrating a recurrent network into firstpass decoding under an efficient approximation.
- MSR SPLAT, a language analysis toolkit40
MSR SPLAT is a toolkit for language analysis that allows easy access to the linguistic analysis tools produced by the NLP group at Microsoft Research, and can be used from a broad set of programming languages.
- BiomedJourney: Counterfactual Biomedical Image Generation by Instruction-Learning from Multimodal Patient Journeys39
This paper presents BiomedJourney, a novel method for counterfactual biomedical image generation by instruction-learning from multimodal patient journeys that substantially outperforms prior state-of-the-art methods in instruction image editing and medical image generation such as InstructPix2Pix and RoentGen.
- ReEval: Automatic Hallucination Evaluation for Retrieval-Augmented Large Language Models via Transferable Adversarial Attacks38
This paper presents ReEval, an LLM-based framework using prompt chaining to perturb the original evidence for generating new test cases for evaluating the LLMs' reliability in using new evidence for answering, and investigates the feasibility of automatically perturbing existing static one for dynamic evaluation.
- The Essential Role of Causality in Foundation World Models for Embodied AI38
It is posited that integrating causal considerations is vital to facilitating meaningful physical interactions with the world and demystify misconceptions about causality in this context and present the outlook for future research.
- Movement Primitive Learning and Generalization: Using Mixture Density Networks37
Representing robot skills as movement primitives that can be learned from human demonstration and adapted to new tasks and situations is a promising approach toward intuitive robot pro gramming.
- Interactive Semantic Parsing for If-Then Recipes via Hierarchical Reinforcement Learning37
This paper develops a hierarchical reinforcement learning (HRL) based agent that significantly improves the parsing performance with minimal questions to the user, and substantially outperforms non-interactive semantic parsers and rule-based agents.
- SITE: Towards Spatial Intelligence Thorough Evaluation35
SITE is introduced, a benchmark dataset towards SI Thorough Evaluation in a standardized format of multi-choice visual question-answering designed to assess large vision-language models' spatial intelligence across diverse visual modalities and SI factors.
- 34
- A study of statistical models for query translation34
Analysis of three statistical query translation models that use different units of translation shows that larger translation units lead to more specific models that usually achieve better translation and cross-language information retrieval results.
- 30
- 29
- A Supervised Learning Approach to Entity Search29
Using a linear model to combine the uses of different features and employing a supervised learning approach in training of the model significantly outperforms the baseline method based solely on co-occurrences in entity search.
- 27
- 27
- 25
- Visually-Augmented Language Modeling23
A novel pre-training framework, named VaLM, to Visually-augment text tokens with retrieved relevant images for Language Modeling, which builds on a novel latent text-image alignment method via an image retrieval module to fetch corresponding images given a textual context.
- Model Tells Itself Where to Attend: Faithfulness Meets Automatic Attention Steering21
AutoPASTA is a method that automatically identifies key contextual information and explicitly highlights it by steering an LLM's attention scores, leading to substantially improved model faithfulness and performance, e.g., an average improvement of 7.95% for LLAMA3-70B-Instruct.
- 21
- 20
- 20
- MagGAN: High-Resolution Face Attribute Editing with Mask-Guided Generative Adversarial Network20
This work presents Mask-guided Generative Adversarial Network (MagGAN) for high-resolution face attribute editing, in which semantic facial masks from a pre-trained face parser are used to guide the fine-grained image editing process.
- Large-scale Expected BLEU Training of Phrase-based Reordering Models20
This work shows how the expected BLEU objective allows us to train a simple linear discriminative reordering model with millions of sparse features on hundreds of thousands of sentences resulting in significant improvements.
- 19
- REO-Relevance, Extraness, Omission: A Fine-grained Evaluation for Image Captioning19
A fine-grained evaluation method REO is presented for automatically measuring the performance of image captioning systems and achieves a higher consistency with human judgments and provides more intuitive evaluation results than alternative metrics.
- 18
- 17
- Visual Imitation Learning of Task-Oriented Object Grasping and Rearrangement16
The Multi-feature Implicit Model (MIMO) is presented, a novel object representation that encodes multiple spatial features between a point and an object in an implicit neural field that embeds the object shapes consistently in different aspects, thus improving its performance in object shape reconstruction from partial observation, shape similarity measure, and modeling spatial relations between objects.
- 16
- Localized Symbolic Knowledge Distillation for Visual Commonsense Models15
This work builds Localized Visual Commonsense models, which allow users to specify (multiple) regions as input and results in more precise VL models of reasoning compared to a baseline of passing a generated referring expression to an LLM.
- Lafite2: Few-shot Text-to-Image Generation14
A novel method for pre-training text-to-image generation model on image-only datasets that considers a retrieval-then-optimization procedure to synthesize pseudo text features and can be applied on different models including generative adversarial networks (GANs) and diffusion models.
- 14
- 13
- 13
- GUI-Libra: Training Native GUI Agents to Reason and Act with Action-aware Supervision and Partially Verifiable RL12
This work presents GUI-Libra, a tailored training recipe that consistently improves both step-wise accuracy and end-to-end task completion and introduces success-adaptive scaling to downweight unreliable negative gradients.
- 12
- Learning a Decision Tree Algorithm with Transformers11
This work introduces MetaTree, a transformer-based model trained via meta-learning to directly produce strong decision trees, which enables MetaTree to emulate these algorithms and intelligently adapt its strategy according to the context, thereby achieving superior generalization performance.
- IMProv: Inpainting-based Multimodal Prompting for Computer Vision Tasks11
IMProv is a generative model that is able to in-context learn visual tasks from multimodal prompts and the empirical results suggest that vision and language prompts are complementary and it is advantageous to use both to achieve better in- context learning performance.
- Training Vision-Language Transformers from Captions11
This work introduces a new model Vision-Language from Captions (VLC) built on top of Masked Auto-Encoders that does not require this supervision and outperforms ViLT on standard benchmarks, and is competitive with many larger models that utilize ROIs trained on annotated bounding-boxes.
- 11
- 11
- Research on Ship Engine Fuel Consumption Prediction Algorithm Based on Adaptive Optimization Generative Network10
A Meta-learning Diffusion Residual Attention Network (MD-RAN) is proposed, which establishes a novel generative modeling approach for marine diesel engine fuel consumption prediction, providing robust support for intelligent engine maintenance and energy efficiency optimization.
- Data Formulator 2: Iterative Creation of Data Visualizations, with AI Transforming Data Along the Way10
With Data Formulator 2, an LLM-powered visualization system to address iterative visualization authoring challenges, users describe their visualization intent with blended UI and natural language inputs, and data transformation are delegated to AI.
- Fast Path Planning for Firefighting UAV Based on A-Star algorithm10
Simulation results showing that the A-Star algorithm can quickly and effectively solve the problem of UAV flight path planning considering positioning errors are shown.
- 9
- NICE: Neural Image Commenting with Empathy9
A novel pre-training model Modeling Affect Generation for Image Comments (MAGIC) which aims to generate comments for images, conditioned on linguistic representations that capture style and affect, and to help generate more empathetic, emotional, engaging and socially appropriate comments is provided.
- 9
- Orchard: An Open-Source Agentic Modeling Framework8
Orchard Env, a lightweight Kubernetes-native environment service that provides reusable primitives for sandbox lifecycle management across task domains, agent harnesses, and training stages, and three agentic modeling recipes are presented, demonstrating that a lightweight, open, harness-agnostic environment layer enables reusable agentic data, training recipes, and evaluation protocols across domains.
- Structural Biases for Improving Transformers on Translation into Morphologically Rich Languages8
This work hypothesizes that structural encoding methods make Transformers more sample-efficient, enabling them to perform better from smaller amounts of data, and investigates two methods for building in such a bias.
- 8
- Towards Understanding Graphical Perception in Large Multimodal Models6
The theory of graphical perception, an approach used to study how humans decode visual information encoded on charts and graphs, is used to develop an evaluation framework for analyzing gaps in LMMs' perception abilities in charts, which underscores several critical limitations of current state-of-the-art LMMs, including GPT-4o.
- 6
- 6
- OpenWebRL: Demystifying Online Multi-turn Reinforcement Learning for Visual Web Agents5
This work introduces OpenWebRL, an open framework for training visual web agents with online multi-turn RL on real websites, and trains OpenWebRL-4B, which establishes a new open-source state of the art on challenging live-web benchmarks.
- 5
- 5
- A comparison of unsupervised methods for Part-of-Speech Tagging in Chinese5
A baseline for unsupervised POS tagging in Chinese is established and some of the strengths and weaknesses of each of the algorithms in POS tagging task are highlighted, to explain the differences based on some preliminary linguistics analysis.
- Improved Monolingual Hypothesis Alignment for Machine Translation System Combination5
The proposed indirect hidden Markov model (IHMM) is proposed to address the synonym matching and word ordering issues in hypothesis alignment and achieved the best Chinese-to-English translation result in the constrained training track of the 2008 NIST Open MT Evaluation.
- Sanity Checks for Agentic Data Science4
A pair of lightweight sanity checks grounded in the Predictability-Computability-Stability (PCS) framework for veridical data science are proposed, characterizing the trustworthiness of each conclusion and finding that in 6 of the datasets an affirmative conclusion is not well-supported, even though a single ADS run may support one.
- 4
- 4
- 4
- 4
- 4
- 3
- 3
- 3
- Test-Time Learning with an Evolving Library2
EvoLib is introduced, a test-time learning framework that enables large language models to accumulate, reuse, and evolve knowledge across problem instances without parameter updates or external supervision, and introduces a principled weighting and consolidation mechanism to support continual improvement.
- 2
- 2
- Online Classification Using a Voted RDA Method2
A voted dual averaging method for on- line classification problems with explicit regularization using the update rule of the regularized dual averaging (RDA) method proposed by Xiao, but only on the subsequence of training examples where a classification error is made.
- Interactive Text Generation1
This work introduces a new Interactive Text Generation task that allows training generation models interactively without the costs of involving real users, by using user simulators that provide edits that guide the model towards a given target text.
- 1
- 1
- 1
- 1
- Ordinal Regression for Information Retrieval1
Results show that ORM significantly outperforms the state-of-the-art language model approaches and OKAPI system in all test sets; and it is more appropriate to view IR as ordinal regression other than binary classification.
- AsgardBench -- Evaluating Visually Grounded Interactive Planning Under Minimal Feedback–
The benchmark zeroes in on a narrower question: can a model actually use what it sees to adapt a plan when things do not go as expected, revealing weaknesses in visual grounding and state tracking that ultimately undermine interactive planning.
- –
- –
- –
- –
- –
- YOLOv7 algorithm and Winograd acceleration circuit for real-time road image recognition and reminder–
- –
- Cascade detector for oriented airplane detection in remote sensing images–
This work innovatively proposes an oriented airplane detector that uses an anchor-free structure to avoid complex calculations related to anchor boxes and significantly reduce the use of memory and time in computing, and replaces the standard convolution with asymmetric convolution.
- –
- –
- –
- NICE: Neural Image Commenting Evaluation with an Emphasis on Emotion and Empathy–
A novel model Modeling Affect Generation for Image Comments (MAGIC) is proposed which aims to generate comments for images, conditioned on linguistic representations that capture style and affect, and to help generate more empathetic, emotional, engaging and socially appropriate comments.
- –
- Reactive Motion Planning for Human-Robot Cooperative Tasks Under Uncertainties–
A Dynamical System-based control architecture with adaptation capabilities is proposed that allows for reactive and adaptive motion planning toward human behavior and shows promising performances to achieve such a complex task in collaboration with human users.
- –
- –
- –
- –
- The structural design, simulation analysis and parameter optimisation of the cheetah robot's leg components–
The methods described herein combines the advantages of CAD and CAE technology, which can significantly enhance not only the accuracy, rationality and speed during the design of the bionic cheetah robot's leg, but also the level of optimisation of biomimetic robots.
- Combining Signals for Cross-Lingual Relevance Feedback–
On an evaluation with web data sampled from a real-world search engine, the proposed cross-lingual feedback model outperforms two state-of-the-art models across two different low-resource languages.
- LMS Adaptive Notch Filter Design Based on Immune Algorithm–
The result of the experiment has shown that the IA can find the optimal learning step of the LMS adaptive notch filter and the convergent speed of algorithm is fast.
Publication data from OpenAlex, with missing venues and authors filled in from Crossref; citation counts are the higher of OpenAlex and Semantic Scholar, last synced 2026-10-12. One-sentence summaries under some papers are written by Semantic Scholar’s model. Citation counts may be lower than on Google Scholar, which indexes more sources.
Report an error
Wrong papers, two people merged into one, or a profile that should not be here? Tell us and we will fix or hide it.