Diffusion models have shown preliminary success in virtual try-on (VTON) task. The typical dual-branch architecture comprises two UNets for implicit garment deformation and synthesized image generation respectively, and has emerged as the recipe for VTON task. Nevertheless, the problem remains challenging to preserve the shape and every detail of the given garment due to the intrinsic stochasticity of diffusion model. To alleviate this issue, we novelly propose to explicitly capitalize on visual correspondence as the prior to tame diffusion process instead of simply feeding the whole garment into UNet as the appearance reference. Specifically, we interpret the fine-grained appearance and texture details as a set of structured semantic points, and match the semantic points rooted in garment to the ones over target person through local flow warping. Such 2D points are then augmented into 3D-aware cues with depth/normal map of target person. The correspondence mimics the way of putting clothing on human body and the 3D-aware cues act as semantic point matching to supervise diffusion model training. A point-focused diffusion loss is further devised to fully take the advantage of semantic point matching. Extensive experiments demonstrate strong garment detail preservation of our approach, evidenced by state-of-the-art VTON performances on both VITON-HD and DressCode datasets. Code is publicly available at: https://github.com/HiDream-ai/SPM-Diff.
LiDAR-based 3D panoptic segmentation often struggles with the inherent sparsity of data from LiDAR sensors, which makes it challenging to accurately recognize distant or small objects. Recently, a few studies have sought to overcome this challenge by integrating LiDAR inputs with camera images, leveraging the rich and dense texture information provided by the latter. While these approaches have shown promising results, they still face challenges, such as misalignment during data augmentation and the reliance on post-processing steps. To address these issues, we propose Image-Assists-LiDAR (IAL), a novel multi-modal 3D panoptic segmentation framework. In IAL, we first introduce a modality-synchronized data augmentation strategy, PieAug, to ensure alignment between LiDAR and image inputs from the start. Next, we adopt a transformer decoder to directly predict panoptic segmentation results. To effectively fuse LiDAR and image features into tokens for the decoder, we design a Geometric-guided Token Fusion (GTF) module. Additionally, we leverage the complementary strengths of each modality as priors for query initialization through a Prior-based Query Generation (PQG) module, enhancing the decoder's ability to generate accurate instance masks. Our IAL framework achieves state-of-the-art performance compared to previous multi-modal 3D panoptic segmentation methods on two widely used benchmarks. Code and models are publicly available at <https://github.com/IMPL-Lab/IAL.git>.
Large language models (LLMs) are often used for infilling tasks, which involve predicting or generating missing information in a given text. These tasks typically require multiple interactions with similar context. To reduce the computation of repeated historical tokens, cross-request key-value (KV) cache reuse, a technique that stores and reuses intermediate computations, has become a crucial method in multi-round interactive services. However, in infilling tasks, the KV cache reuse is often hindered by the structure of the prompt format, which typically consists of a prefix and suffix relative to the insertion point. Specifically, the KV cache of the prefix or suffix part is frequently invalidated as the other part (suffix or prefix) is incrementally generated. To address the issue, we propose EFIM, a transformed prompt format of FIM to unleash the performance potential of KV cache reuse. Although the transformed prompt can solve the inefficiency, it exposes subtoken generation problems in current LLMs, where they have difficulty generating partial words accurately. Therefore, we introduce a fragment tokenization training method which splits text into multiple fragments before tokenization during data processing. Experiments on two representative LLMs show that LLM serving with EFIM can lower the latency by 52% and improve the throughput by 98% while maintaining the original infilling capability.EFIM's source code is publicly available at https://github.com/gty111/EFIM.
Large Language Models (LLMs) possess an extraordinary capability to produce text that is not only coherent and contextually relevant but also strikingly similar to human writing. They adapt to various styles and genres, producing content that is both grammatically correct and semantically meaningful. Recently, LLMs have been misused to create highly realistic phishing emails, spread fake news, generate code to automate cyber crime, and write fraudulent scientific articles. Additionally, in many real-world applications, the generated content including style and topic and the generator model are not known beforehand. The increasing prevalence and sophistication of artificial intelligence (AI)-generated texts have made their detection progressively more challenging. Various attempts have been made to distinguish machine-generated text from human-authored content using linguistic, statistical, machine learning, and ensemble-based approaches. This work focuses on two primary objectives Task-A, which involves distinguishing human-written text from machine-generated text, and Task-B, which attempts to identify the specific LLM model responsible for the generation. Both of these tasks are based on fine tuning of Generative Pre-trained Transformer (GPT_4o-mini), Large Language Model Meta AI (LLaMA) 3 8B, and Bidirectional Encoder Representations from Transformers (BERT). The fine-tuned version of GPT_4o-mini and the BERT model has achieved accuracies of 0.9547 for Task-A and 0.4698 for Task-B.
대조적 예측 코딩(CPC)을 3GPP 규격 준수 CSI 압축 아키텍처에 통합한 통합 압축-예측 프레임워크를 제안하며, 고차원 CSI 행렬 대신 잠재 표현을 예측하고 1-SGCS와 InfoNCE 결합 목적 함수로 재구성 정확도와 시간적 예측 일관성을 공동 최적화한다. Nokia, Oppo, CATT의 3GPP 규격 데이터셋 평가에서 CPC-before-Compression은 기준선 대비 디코더 GFLOPs를 32배 낮추면서 90% 이상의 재구성 정확도를 달성하고, CPC-after-Compression은 동일한 인코더 풋프린트와 64비트 피드백 오버헤드를 유지한다.
에이전트 능력은 모델만으로 결정되지 않는다. 메모리 관리, 계획 전략, 행동 프로토콜, 도구/스킬 오케스트레이션을 포괄하는 에이전트 하네스(agent harness)가 기반 파운데이션 모델의 기여를 압도할 수 있다. 그러나 하네스 설계는 여전히 수동적이고 작업별이며 근본적으로 확장 불가능하다. 우리는 임의의 상용 에이전트형 LLM에 대해 과제에 적응하는 에이전트 하네스를 즉석에서 합성하도록 학습된 하네스 지능(harness intelligence) 모델인 JIT-Agent를 제시한다. 우리는 에이전트 하네스를 고정된 4모듈 프로토콜로 규율되는 조합 가능하고 기계 생성 가능한 산출물로 형식화하고, JIT-Agent가 당면 과제에 맞게 하네스를 맞춤화하고, 안정적이고 신뢰할 수 있는 실행을 위해 하네스를 수리하며, 이전 하네스 구성의 확장되는 아카이브로부터 성능 신호를 지식 증류(distillation)하여 자기 진화하도록 학습한다. JIT-Agent를 하네스 헬퍼로 갖춘 DeepSeek-V4-Flash는 DeepSearchQA에서 GPT-5.6을 능가(+9.1)하고 OdysseyBench에서(+4.3) 능가하며, 이미 강력한 GLM-5.2는 최대 +20.2점 향상을 얻는다. 통제된 평가에서 JIT-Agent가 생성한 하네스는 OpenCode 및 Claude Code와 같은 성숙한 에이전트 런타임과 성능 면에서 경쟁력 있으며 DeepSeek V4, Mimo-V2.5, Qwen3.6의 다규모 모델군을 일관되게 개선한다. 우리가 아는 한 JIT-Agent는 적시 하네스 생성을 위해 목적에 맞게 구축된 최초의 모델이며, 하네스 지능을 모델 스케일링과 직교하는 학습 가능하고 전이 가능하며 누적되는 에이전트 능력의 차원으로 확립한다.
키-값(KV) 캐시는 장문맥 대규모 언어 모델(LLM) 추론에서 지배적인 메모리 병목이며, 문맥 길이에 선형적으로 증가한다. 분수 비트 격자에서의 균일 KV 양자화는 완만하게 저하되지 않음을 보인다. 사전 지정된 다중 시드 통계 프로토콜 하에서, affine 양자화기를 사용한 Llama-3.1-8B-Instruct는 값당 2.322 code bits까지 FP16 KV와 통계적으로 구별 불가능하며 2.0 bits에서 붕괴한다. 이는 (2.0, 2.322] 구간의 품질 절벽(quality cliff)으로, 생성 시점 양자화와 다중 턴 대화에서도 재현되며 Mistral-7B로 전이된다. 이 절벽은 중요도 인식 혼합 정밀도를 재구성한다. 절벽 위에서는 여덟 가지 모델 내부 중요도 지표가 통계적으로 상호 교환 가능하므로, 혼합의 이점은 격자 보간에 있으며 균일 양자화가 실현할 수 없는 평균 정밀도에 도달한다. SemKV는 모든 토큰을 보존하고, 모델 내부 점수로 토큰 순위를 매기며, 절벽 위의 두 인접 정밀도를 할당한다. 이를 통해 전체 KV 대비 통계적으로 검출 가능한 품질 차이 없이(n=900, 세 시드) 측정된 6.0배 저장 공간 절감을 달성하고, 1.5배 더 큰 메모리 예산을 부여한 FP16 토큰 가지치기를 능가한다. affine 기반을 왜곡 최적화 양자화기(TurboQuant-MSE)로 교체하면 테스트한 모든 프로토콜에서 절벽이 낮아져, 손실 미검출 운영 지점을 7.9배까지 끌어올린다. 요약하면, 대상 배포 설정에서 절벽을 측정한 뒤 그 위에서 보간하면 된다.
최근 대규모 언어 모델(LLM) 기반 text-to-SQL 시스템은 표준 벤치마크에서 인상적인 성능을 달성하지만, 사용자 질의가 스키마에 담기지 않고 자연어 질문에도 명시되지 않은 비즈니스 로직, 데이터 관례, 분석 관행 같은 도메인 특화 지식에 암묵적으로 의존할 때는 어려움을 겪는다. 과거 SQL 쿼리 로그는 이러한 지식의 유용한 출처이나, 기존 벤치마크는 이력 기반 접근법의 평가를 충분히 지원하지 않는다. 이 공백을 메우기 위해 우리는 11개 데이터베이스에 걸친 1,393개 과제로 구성된 벤치마크 BIRD-History를 제안하며, 과거 SQL 스크립트를 이용해 명시가 부족한 자연어 질문을 그라운딩(grounding)하는 text-to-SQL 시스템 능력을 평가하도록 설계했다. 각 과제에는 어떤 과거 쿼리에 관련 지식이 담겨 있는지, 어떤 SQL 절이 그 지식을 인코딩하는지를 지정하는 정답 레이블이 주석으로 달려 있어, 검색 효과와 지식 활용을 체계적으로 평가할 수 있다. 벤치마크와 함께 우리는 과거 SQL 스크립트에서 다섯 가지 유형의 외부 지식을 추출한 뒤, 쿼리 생성을 위해 관련 조각을 검색·재순위화하는 플러그인 검색기를 제안한다. 이 검색기는 프롬프트 수정 없이 기존 퓨샷(few-shot) text-to-SQL 파이프라인에 원활히 통합된다. 실험은 네 가지 text-to-SQL 시스템에서 일관된 개선을 보여 주며, 명시가 부족한 질의 처리에 과거 쿼리 로그를 활용하는 가치를 부각한다. 데이터셋과 코드는 https://github.com/zjuidg/BIRD-History 에서 오픈소스로 공개한다.
메시지 패싱(message passing)은 그래프 신경망(GNN)의 핵심 메커니즘으로, 이웃 노드의 정보를 집계하여 노드 임베딩을 반복적으로 갱신할 수 있게 한다. 그래프 합성곱 네트워크(GCN)는 합성곱 연산을 그래프 구조에 맞게 적용함으로써 인접 노드의 특징을 효과적으로 결합할 수 있게 하는 대표적인 접근법이다. 그러나 GCN은 복잡하거나 동적인 데이터에서 어려움을 겪는다. 장거리 의존성(long-range dependency)을 포착하려면 흔히 더 깊은 레이어가 필요한데, 이는 계산 비용을 증가시킬 뿐만 아니라 노드 임베딩이 서로 구별되지 않게 되는 과평활화(over-smoothing)를 초래한다. 이러한 문제를 극복하기 위해, 광범위한 매개변수 튜닝 없이도 반복적인 메시지 패싱 동역학을 활용하여 안정적인 정보 전파를 가능하게 하는 리저버 컴퓨팅(reservoir computing)이 GNN에 통합되어 왔다. 그러나 이러한 가능성에도 불구하고, 기존의 리저버 기반 모델은 구조화된 합성곱 메커니즘이 부족하여 다중 홉(multi-hop) 이웃 정보를 정확하게 집계하는 능력이 제한된다. 이러한 한계를 해결하기 위해, 본 논문에서는 리저버 동역학과 구조화된 그래프 합성곱을 통합한 RGC-Net(Reservoir-based Graph Convolutional Network)을 제안한다. 주요 기여는 다음과 같다. (i) 고정된 무작위 리저버 가중치와 누설 적분기(leaky integrator)를 갖추어 특징 보존을 강화한 새로운 합성곱 프레임워크, (ii) 그래프 분류를 위한 견고하고 적응력 있는 모델, (iii) 동적 뇌 연결성(dynamic brain connectivity)에 적용할 수 있는, RGC-Net 기반의 그래프 생성용 트랜스포머. 광범위한 실험 결과, RGC-Net은 뇌 그래프 진화(brain graph evolution)를 포함한 분류 및 생성 과제에서 최고 수준(state-of-the-art)의 성능을 달성했으며, 더 빠른 수렴과 완화된 과평활화를 보였다. 소스 코드는 https://github.com/basiralab/RGC-Net 에서 확인할 수 있다.
정보 도출(information elicitation)은 상호작용이 잠재적으로 가치 있는 다양한 방향으로 전개될 수 있는 개방형 정보 탐색 문제로, 정보 도출자(elicitor)는 새로운 정보가 드러남에 따라 어떤 정보를 추구해야 할지 지속적으로 결정해야 합니다. 에이전트 기반 정보 도출에서는 이러한 결정을 파운데이션 모델에 위임할 수 있으나, 모델 선택이 결과적인 정보 탐색 행동을 어떻게 형성하는지는 아직 충분히 연구되지 않았습니다. 본 연구에서는 정보 가치에 대한 판단이 대규모 언어 모델(LLM) 전반에서 어떻게 달라지며, 이러한 차이가 순차적 정보 탐색을 어떻게 형성하는지 살펴봅니다. 먼저 여러 모델 군과 매개변수(parameter) 규모에 걸친 11개 LLM을 대상으로, 공유된 정보 세트와 도출 목표를 사용하여 이러한 판단을 조사합니다. 그런 다음 서로 다른 모델들이 동일한 정보 공간을 접하고 동일한 선택 규칙을 사용하도록 하는 통제된 정보 도출 시뮬레이션을 개발하여, 질문 생성 및 응답자 행동과 이러한 판단을 분리합니다. 이 설정을 사용하여 정보 도출 과정 전반에서 모델 고유의 정보 탐색 선호도로부터 나타나는 폭-깊이(breadth-depth) 행동 특성을 규명합니다. 나아가 상호작용 이력이 유망한 정보의 평가와 후속 선택을 어떻게 변화시키는지 살펴봅니다. 정보 도출자에게 주어지는 기회, 정보 탐색 선호도를 조작화하는 데 사용된 응답 라벨, 상호작용 이력의 유무, 중복성이 평가에 명시적으로 관련되는지 여부에 대한 민감도 분석 및 절제 연구(ablation)를 통해 이러한 발견의 견고성과 한계를 검증합니다. 프로젝트 코드, 데이터, 궤적 파일은 https://github.com/infosenselab/open-elicitation 에서 확인할 수 있습니다.
개인화된 비디오 생성 기술의 급격한 발전으로 몇 개의 인물 얼굴 비디오만으로도 합성 비디오에서 해당 신원을 재현할 수 있게 되면서, 신원 프라이버시에 대한 심각한 우려가 제기되고 있습니다. 본 보고서에서는 제한적이고 감지할 수 없는 섭동(perturbation) 예산 내에서 무단 맞춤 제작(customization)으로부터 얼굴 비디오를 보호하는 것을 목표로 하는 학습 불가능한 비디오(Unlearnable Videos) 대상 ECCV 2026 LifeGenIP 챌린지의 두 단계 모두에서 1위를 차지한 솔루션을 제시합니다. 1단계(Phase 1)에서는 후속 비디오 표현이 참조 조건에 계속 의존하도록 유지하면서도 참조 조건이 전달하는 정보를 저하시키는 한 쌍의 상호보완적 손실인 참조 용량 붕괴(Reference Capacity Collapse)와 함께, 미지의 추론(inference) 조건 전반에서 강건성을 향상시키는 모션 유도 EOT(Motion-Guided EOT) 메커니즘을 도입합니다. 2단계(Phase 2)에서는 튜닝 기반 맞춤 제작과 미확인 생성 모델까지 추가로 고려합니다. 단일 섭동 내에서 두 가지 위협을 모두 해결하기 위해, 얼굴과 배경 영역을 분리하여 보호하는 2층(two-layer) 프레임워크를 설계하였으며, 전자에서는 합성 타깃 신원으로부터 구조 정렬 섭동을 구성하고 후자에서는 1단계의 화이트박스 교란을 강화합니다. 제안하는 방법은 화이트박스 참조 기반 맞춤 제작 환경에서 거의 완벽한 보호를 달성하며, 다른 맞춤 제작 패러다임, 다양한 모델 및 시간적 공격(temporal attack) 하에서도 여전히 효과적입니다. 두 솔루션 모두 최근 연구인 LETHE를 기반으로 구축되었습니다. 본 솔루션이 무단 맞춤 제작으로부터 얼굴 비디오를 보호하는 미래 연구에 영감을 주기를 기대합니다. 코드는 https://github.com/AntigoneRandy/KVeil 에서 이용할 수 있습니다.
Every natural text is written in some style. Style is formed by a complex combination of different stylistic factors, including formality markers, emotions, metaphors, etc. One cannot form a complete understanding of a text without considering these factors. The factors combine and co-vary in complex ways to form styles. Studying the nature of the co-varying combinations sheds light on stylistic language in general, sometimes called cross-style language understanding. This paper provides the benchmark corpus (xSLUE) that combines existing datasets and collects a new one for sentence-level cross-style language understanding and evaluation. The benchmark contains text in 15 different styles under the proposed four theoretical groupings: figurative, personal, affective, and interpersonal groups. For valid evaluation, we collect an additional diagnostic set by annotating all 15 styles on the same text. Using xSLUE, we propose three interesting cross-style applications in classification, correlation, and generation. First, our proposed cross-style classifier trained with multiple styles together helps improve overall classification performance against individually-trained style classifiers. Second, our study shows that some styles are highly dependent on each other in human-written text. Finally, we find that combinations of some contradictive styles likely generate stylistically less appropriate text. We believe our benchmark and case studies help explore interesting future directions for cross-style research. The preprocessed datasets and code are publicly available.
Interpreting objects with basic geometric primitives has long been studied in computer vision. Among geometric primitives, superquadrics are well known for their ability to represent a wide range of shapes with few parameters. However, as the first and foremost step, recovering superquadrics accurately and robustly from 3D data still remains challenging. The existing methods are subject to local optima and sensitive to noise and outliers in real-world scenarios, resulting in frequent failure in capturing geometric shapes. In this paper, we propose the first probabilistic method to recover superquadrics from point clouds. Our method builds a Gaussian-uniform mixture model (GUM) on the parametric surface of a superquadric, which explicitly models the generation of outliers and noise. The superquadric recovery is formulated as a Maximum Likelihood Estimation (MLE) problem. We propose an algorithm, Expectation, Maximization, and Switching (EMS), to solve this problem, where: (1) outliers are predicted from the posterior perspective; (2) the superquadric parameter is optimized by the trust-region reflective algorithm; and (3) local optima are avoided by globally searching and switching among parameters encoding similar superquadrics. We show that our method can be extended to the multi-superquadrics recovery for complex objects. The proposed method outperforms the state-of-the-art in terms of accuracy, efficiency, and robustness on both synthetic and real-world datasets. The code is at http://github.com/bmlklwx/EMS-superquadric_fitting.git.
Effectiveness and interpretability are two essential properties for trustworthy AI systems. Most recent studies in visual reasoning are dedicated to improving the accuracy of predicted answers, and less attention is paid to explaining the rationales behind the decisions. As a result, they commonly take advantage of spurious biases instead of actually reasoning on the visual-textual data, and have yet developed the capability to explain their decision making by considering key information from both modalities. This paper aims to close the gap from three distinct perspectives: first, we define a new type of multi-modal explanations that explain the decisions by progressively traversing the reasoning process and grounding keywords in the images. We develop a functional program to sequentially execute different reasoning steps and construct a new dataset with 1,040,830 multi-modal explanations. Second, we identify the critical need to tightly couple important components across the visual and textual modalities for explaining the decisions, and propose a novel explanation generation method that explicitly models the pairwise correspondence between words and regions of interest. It improves the visual grounding capability by a considerable margin, resulting in enhanced interpretability and reasoning performance. Finally, with our new data and method, we perform extensive analyses to study the effectiveness of our explanation under different settings, including multi-task learning and transfer learning. Our code and data are available at https://github.com/szzexpoi/rex.
Co-salient object detection, with the target of detecting co-existed salient objects among a group of images, is gaining popularity. Recent works use the attention mechanism or extra information to aggregate common co-salient features, leading to incomplete even incorrect responses for target objects. In this paper, we aim to mine comprehensive co-salient features with democracy and reduce background interference without introducing any extra information. To achieve this, we design a democratic prototype generation module to generate democratic response maps, covering sufficient co-salient regions and thereby involving more shared attributes of co-salient objects. Then a comprehensive prototype based on the response maps can be generated as a guide for final prediction. To suppress the noisy background information in the prototype, we propose a self-contrastive learning module, where both positive and negative pairs are formed without relying on additional classification information. Besides, we also design a democratic feature enhancement module to further strengthen the co-salient features by readjusting attention values. Extensive experiments show that our model obtains better performance than previous state-of-the-art methods, especially on challenging real-world cases (e.g., for CoCA, we obtain a gain of 2.0% for MAE, 5.4% for maximum F-measure, 2.3% for maximum E-measure, and 3.7% for S-measure) under the same settings. Code will be released soon.
Visual grounding, i.e., localizing objects in images according to natural language queries, is an important topic in visual language understanding. The most effective approaches for this task are based on deep learning, which generally require expensive manually labeled image-query or patch-query pairs. To eliminate the heavy dependence on human annotations, we present a novel method, named Pseudo-Q, to automatically generate pseudo language queries for supervised training. Our method leverages an off-the-shelf object detector to identify visual objects from unlabeled images, and then language queries for these objects are obtained in an unsupervised fashion with a pseudo-query generation module. Then, we design a task-related query prompt module to specifically tailor generated pseudo language queries for visual grounding tasks. Further, in order to fully capture the contextual relationships between images and language queries, we develop a visual-language model equipped with multi-level cross-modality attention mechanism. Extensive experimental results demonstrate that our method has two notable benefits: (1) it can reduce human annotation costs significantly, e.g., 31% on RefCOCO without degrading original model's performance under the fully supervised setting, and (2) without bells and whistles, it achieves superior or comparable performance compared to state-of-the-art weakly-supervised visual grounding methods on all the five datasets we have experimented. Code is available at https://github.com/LeapLabTHU/Pseudo-Q.
5G is the 5th generation cellular network protocol. It is the state-of-the-art global wireless standard that enables an advanced kind of network designed to connect virtually everyone and everything with increased speed and reduced latency. Therefore, its development, analysis, and security are critical. However, all approaches to the 5G protocol development and security analysis, e.g., property extraction, protocol summarization, and semantic analysis of the protocol specifications and implementations are completely manual. To reduce such manual effort, in this paper, we curate SPEC5G the first-ever public 5G dataset for NLP research. The dataset contains 3,547,586 sentences with 134M words, from 13094 cellular network specifications and 13 online websites. By leveraging large-scale pre-trained language models that have achieved state-of-the-art results on NLP tasks, we use this dataset for security-related text classification and summarization. Security-related text classification can be used to extract relevant security-related properties for protocol testing. On the other hand, summarization can help developers and practitioners understand the high level of the protocol, which is itself a daunting task. Our results show the value of our 5G-centric dataset in 5G protocol analysis automation. We believe that SPEC5G will enable a new research direction into automatic analyses for the 5G cellular network protocol and numerous related downstream tasks. Our data and code are publicly available.
The burgeoning field of camouflaged object detection (COD) seeks to identify objects that blend into their surroundings. Despite the impressive performance of recent models, we have identified a limitation in their robustness, where existing methods may misclassify salient objects as camouflaged ones, despite these two characteristics being contradictory. This limitation may stem from lacking multi-pattern training images, leading to less saliency robustness. To address this issue, we introduce CamDiff, a novel approach inspired by AI-Generated Content (AIGC) that overcomes the scarcity of multi-pattern training images. Specifically, we leverage the latent diffusion model to synthesize salient objects in camouflaged scenes, while using the zero-shot image classification ability of the Contrastive Language-Image Pre-training (CLIP) model to prevent synthesis failures and ensure the synthesized object aligns with the input prompt. Consequently, the synthesized image retains its original camouflage label while incorporating salient objects, yielding camouflage samples with richer characteristics. The results of user studies show that the salient objects in the scenes synthesized by our framework attract the user's attention more; thus, such samples pose a greater challenge to the existing COD models. Our approach enables flexible editing and efficient large-scale dataset generation at a low cost. It significantly enhances COD baselines' training and testing phases, emphasizing robustness across diverse domains. Our newly-generated datasets and source code are available at https://github.com/drlxj/CamDiff.
In the past years, AI has seen many advances in the field of NLP. This has led to the emergence of LLMs, such as the now famous GPT-3.5, which revolutionise the way humans can access or generate content. Current studies on LLM-based generative tools are mainly interested in the performance of such tools in generating relevant content (code, text or image). However, ethical concerns related to the design and use of generative tools seem to be growing, impacting the public acceptability for specific tasks. This paper presents a questionnaire survey to identify the intention to use generative tools by employees of an IT company in the context of their work. This survey is based on empirical models measuring intention to use (TAM by Davis, 1989, and UTAUT2 by Venkatesh and al., 2008). Our results indicate a rather average acceptability of generative tools, although the more useful the tool is perceived to be, the higher the intention to use seems to be. Furthermore, our analyses suggest that the frequency of use of generative tools is likely to be a key factor in understanding how employees perceive these tools in the context of their work. Following on from this work, we plan to investigate the nature of the requests that may be made to these tools by specific audiences.
Spiking Neural Networks (SNNs) are recognized as the candidate for the next-generation neural networks due to their bio-plausibility and energy efficiency. Recently, researchers have demonstrated that SNNs are able to achieve nearly state-of-the-art performance in image recognition tasks using surrogate gradient training. However, some essential questions exist pertaining to SNNs that are little studied: Do SNNs trained with surrogate gradient learn different representations from traditional Artificial Neural Networks (ANNs)? Does the time dimension in SNNs provide unique representation power? In this paper, we aim to answer these questions by conducting a representation similarity analysis between SNNs and ANNs using Centered Kernel Alignment (CKA). We start by analyzing the spatial dimension of the networks, including both the width and the depth. Furthermore, our analysis of residual connections shows that SNNs learn a periodic pattern, which rectifies the representations in SNNs to be ANN-like. We additionally investigate the effect of the time dimension on SNN representation, finding that deeper layers encourage more dynamics along the time dimension. We also investigate the impact of input data such as event-stream data and adversarial attacks. Our work uncovers a host of new findings of representations in SNNs. We hope this work will inspire future research to fully comprehend the representation power of SNNs. Code is released at https://github.com/Intelligent-Computing-Lab-Yale/SNNCKA.