Artificial Intelligence news — Tuesday 27 October 2026

AI Times (Korea) 23:04 UTC Translated from Korean

On 26th (local time), NYT was represented by the “nearly everything is at stake”. Antropic’s listing was a test bench that validates the growth view and investment hot air across the AI industry, without compromising in a single…

AI Times (Korea) 23:04 UTC Translated from Korean

Google has released the next-generation world model ‘Genie 4’, which creates a virtual world in real time, and creates a dialogue to the movements and conversations of the emerging people. We have developed physical phenomenon and human…

Google DeepMind 16:01 UTC

Genie 4: our new world model unlocking unprecedented realism in real-time simulation.

Towards Data Science 15:30 UTC

I replayed correct model decisions through four deliberately broken browser actions, and a simple page check missed half of the wrong clicks, which shows that noticing a fault and recovering from it are different skills. The post An AI…

AI Business 14:24 UTC

The rising use of AI-related terminology reflects a shift in enterprise employees' perceptions and the growing integration of AI into their daily work.

Towards Data Science 14:00 UTC

A separation of powers for control of autonomous agents The post Who Watches the AI? appeared first on Towards Data Science .

Towards Data Science 12:30 UTC

A framework for staying relevant in the age of artificial intelligence The post In the Age of Borrowed Thinking, How To Become More Valuable as AI Gets Smarter appeared first on Towards Data Science .

MIT Technology Review Biotech 12:10 UTC

This is today’s edition of The Download , our weekday newsletter that provides a daily dose of what’s going on in the world of technology. Will the real Anthropic please stand up? After weeks of frenzied speculation, Anthropic has filed…

MIT Technology Review Biotech 11:19 UTC

Every manufacturer has its own industrial intelligence. Some is documented, and some resides in the knowledge of experienced workers. Consider an experienced operator preparing to restart a line after a changeover. The machine settings…

Towards Data Science 11:00 UTC

And that’s not even the worrying part. The post Developers Felt 20% Faster With AI. They Were 19% Slower. appeared first on Towards Data Science .

AI Times (Korea) 10:08 UTC Translated from Korean

NAVER, NVIDIA, Brookfield 3, is a global AI infrastructure project that promotes 1 gigawatt (GW). NAVER (CEO Choi Su-yeon) signed an agreement to secure computing infrastructure of Brookfield and $90 billion (about 12 trillion KRW), and…

AI Times (Korea) 10:08 UTC Translated from Korean

China Nanbeige LLM Institute published an agent-type AI model ‘Nanbeige4.3-33B’ that performs coding and working automation. With an open source model with a 3 billion parameters, the Institute’s self-established coding and office…

AI Times (Korea) 10:08 UTC Translated from Korean

Even though it’s complicated to configure the working environment of AI coding agent, the code creation accuracy is not much higher, but the cost of the future has been increased by up to 3 times. This is an analysis that reduces…

AI Times (Korea) 09:07 UTC Translated from Korean

The exhibition is held in a place that showcases optical convergence technologies such as AI semiconductor, optical communication, Quantum, Security, and Future Mobility. The 24th PHOTONICS KOREA will be held at the Daejung Convention…

AI Times (Korea) 09:07 UTC Translated from Korean

AI-produced video surveillance enhances coastal low-land security checks and disaster prevention checks against the 28th seawater water level rise. Thumbong said that the seawater water level from 1 pm to 5 pm will rise at 4.92m or higher…

AI Times (Korea) 09:07 UTC Translated from Korean

American online dictionary.com selected ‘Maxxing’ as a word for the 2026 year. We’ve introduced ‘Tokenmaxxing’, which maximizes AI usage in the workplace. The company culture is also reflected in the language trends of this year. Dick…

AI Times (Korea) 09:07 UTC Translated from Korean

■ Javis&Villon (CEO Kim Bum-Soon) announced that it has built a self-platform ‘carter’ that integrates an in-house AI agent to save up to 45% of AI operations. MORN LASER teams always stand behind all customers and products. We will give…

AI Times (Korea) 09:07 UTC Translated from Korean

Antropic Founder said that 7 people will not introduce advertisements to ‘Call’. As we will focus on providing useful answers to users than advertising revenues, we will be pooled with the strategy to differentiate with openAI that…

AI Times (Korea) 09:07 UTC Translated from Korean

Chips&Media (CEO Kim Sang-hyun) announced that it recorded KRW 71.6 billion in sales (connected basis) and KRW 16.6 billion in sales profit in the new license supply in AI SoC and Robotics. 8.2%, 8.7%, respectively The licensing revenue…

AI Times (Korea) 09:07 UTC Translated from Korean

In the construction process of Gwangju Metropolitan City Railway Line No. 2, the installation was completed, and the fact that the construction documents were written in the condition that some construction did not end. The construction…

AI Times (Korea) 09:07 UTC Translated from Korean

We have been revised to ensure that our employees are working on a clean workplace. We have been revised by industrial safety measures to ensure that the employees are working directly at the office. The Company shall notify the Company…

AI Times (Korea) 09:07 UTC Translated from Korean

The value of the asset was confirmed that the value of the value of the time is dropped, and the concept of the imagination is not true in the GPU market. U.S. venture capital &risson Horus (a16z) published five major GPU rental prices…

AI Times (Korea) 09:07 UTC Translated from Korean

Immigrants 2 years old living in South Korea was inherited by 2 million people. The total population is 1.12%, the highest among 17 attempts in the country. The result of analyzing the material by attempting the ‘2025 I-Jobyeong-in-gu…

AI Times (Korea) 09:00 UTC Translated from Korean

The LangGraph agent of Tedinot is a non-legal non-legal non-legal non-legal non-legal Non-legal Non-legal Non-legal Non-legal Non-legal Non-legal Non-legal Non-legal Non-legal Non-legal Non-legal Non-legal Non-legal Non-legal Non-legal…

MIT Technology Review Biotech 09:00 UTC

Transformers are a crucial but nearly invisible part of our grid. The devices help efficiently move power across long distances, but they’re complicated and in short supply. And given that their basic design hasn’t changed much for about a…

AI Times (Korea) 08:05 UTC Translated from Korean

Apple released a new iPad mini with organic display and next generation 'A20 Pro' chip. We improved the design, such as reducing the bezel, raising the screen size of 8.5 inches and reducing the thickness and weight, according to Apple, AI…

AI Times (Korea) 08:05 UTC Translated from Korean

Alababa’s ‘Quyuan’ has been revealed that the information of the running environment during cybersecurity testing and the fact that the files were sent to Alibaba Cloud servers in Beijing, China. The model will not use external tools in…

AI Times (Korea) 08:05 UTC Translated from Korean

Cadence Design Systems, a Semiconductor Design Software company, published a “Cadence Maestro”, a design-specific AI model suite for semiconductor manufacturers to run their data centers. It is explained that it can increase the quality of…

AI Times (Korea) 08:05 UTC Translated from Korean

The Korean bank has a boundary on the structure of the fundraising of the AI industry. As an external asset dependence and circulation finance grows, the AI risk boundary is expanding, and the Korean connected with global supply chain is…

AI Times (Korea) 08:01 UTC Translated from Korean

OpenAI has released ‘GPT-6.1 terra’ which increases computer operation performance and lowers usage cost. We aimed at the company and developer demands to automate repeat operations such as document processing and coding. The API price is…

AI Times (Korea) 07:04 UTC Translated from Korean

Meta will propose a stock issue of up to $9 billion (about $126 trillion) to secure funds required for AI infrastructure expansion. As an investment in AI data center and computing infrastructure, BigTech companies are actively utilizing…

AI Times (Korea) 07:04 UTC Translated from Korean

SpaceX signed an AI chip and server equipment financial lease agreement with Apollo Global Management and $400 billion (about $53 trillion). On the day of the day, the CEO of Sunon Musk has completed the funding for buying “GPU” and the…

AI Times (Korea) 07:04 UTC Translated from Korean

Bank of America, Citi Group, Morgan Stanley, etc., is a large bank and global alternative investment management company Blackstone to build an AI computing infrastructure of Ann Tropic. Broadcom’s AI chip and networking equipment are sold…

AI Times (Korea) 07:04 UTC Translated from Korean

Apple began the Korean service of the next generation AI book ‘Siri AI’. Apple has been expanding Siri AI into 5 languages, including Korean, Korean, and so on with the iOS 27.2 software update released on 26 (local time). Last September…

AI Times (Korea) 07:04 UTC Translated from Korean

KT launches ‘KT Token Factory’ service for companies who can choose and manage and pay 500 or more AI models at one platform. KT announced that it will launch the KT token factory on 30 days and will develop as a new growth business…

OpenAI Blog 07:00 UTC

OpenAI and Oracle are expanding their partnership to bring business applications and AI agents into ChatGPT and power new Oracle offerings.

AI Times (Korea) 07:00 UTC Translated from Korean

The U.S. large investment firm Apollo Research analyzes the AI financial performance of the U.S. companies, except BigTech, and the U.S. media are in line with this. The 25th (local time) Bloomberg was the title of “Apollow Economist, the…

AI Times (Korea) 06:28 UTC Translated from Korean

Antropicy, which promotes IPO, has published a risk factor that threatens business such as AI security accidents and data center relapse, and semiconductor supply chain reliance. The next-generation model learning slows for more than a…

AI Times (Korea) 06:15 UTC Translated from Korean

The company’s IPO has revealed its long-term expenditures and high dependence on BigTech as well as its listing application. Over the next decade, the expenditure exceeds $600 billion (approx. 796 trillion KRW), and the equivalent amount…

AI Times (Korea) 06:02 UTC Translated from Korean

Apple’s first folderable smartphone ‘iPhone Duo’ has accidentally accidental consumer complaints with the problem that the internal screen is broken or the white line appears. Some users claimed that the screen was broken in just a few…

AI Times (Korea) 06:02 UTC Translated from Korean

HyperExcel (CEO Kim Jo-young) announced that the director of Samsung Securities Capital Market, Kim Joon-yeon, Kim Jo-young was appointed as Chief Financial Officer (CFO). We are planning to strengthen overseas investment attracting and…

AI Times (Korea) 06:02 UTC Translated from Korean

Cathedral (Cathedral), a cybersecurity startup, founded by greetings from the U.S. Government Efficiency Department (DOGE), found a path that can penetrate the internal system of Andropic. We have found a vulnerability to access sensitive…

AI Times (Korea) 05:52 UTC Translated from Korean

Antropic has released a listing application and entered a full-fledged procedure for Nasdang. As a company’s AI demand, the first half of the year’s sales increased by about 14 times in the same period last year, stood at ensuring…

AI Times (Korea) 05:01 UTC Translated from Korean

OpenAI announced that it was created with AI to prove the development of the field of computer science and theory. A study that proves long speculations or reversing the existing hypothesis, the external experts will review the proof and…

AI Times (Korea) 04:28 UTC Translated from Korean

The third week of AI Times and East Eide (October 16 to 22) issue analysis results include major political issues such as DPRK and DPRK, which include major political issues such as DPRK and reverse osmosis fires, and other events and…

AI Times (Korea) 04:00 UTC Translated from Korean

Microsoft (MS) has launched a voluntary retirement program for more than 6 million employees worldwide. In order to increase the cost efficiency by adjusting the workforce structure according to a rapid AI infrastructure investment. MS…

AI Times (Korea) 04:00 UTC Translated from Korean

AI Infrastructure Professional Rumbleup (CEO) will be listed on the KOSDAQ market on November 3rd. It was established as KRW 40,000 (KRW 414 billion in total). In the case of the investor’s target demand, the total 1129 institutions were…

arXiv cs.CV 04:00 UTC

arXiv:2606.31179v2 Announce Type: replace-cross Abstract: As AI agents become increasingly capable of complex, long-horizon reasoning, rigorous evaluation is essential for measuring progress toward real-world healthcare applications…

arXiv cs.CL 04:00 UTC

arXiv:2602.02518v3 Announce Type: replace-cross Abstract: Large language models (LLMs) increasingly rely on external knowledge to improve factuality, yet many knowledge sources on the Web, such as scholarly indexes and product catalogs…

arXiv cs.CV 04:00 UTC

arXiv:2610.33868v1 Announce Type: new Abstract: Language-aware 3D Gaussian Splatting (3DGS) has advanced language-driven 3D scene understanding by incorporating semantic and relational cues into reconstructed scenes. However…

arXiv cs.CV 04:00 UTC

arXiv:2610.34734v1 Announce Type: new Abstract: Frequency-domain information is beneficial for semantic segmentation. However, existing frequency-based methods often suffer from insufficient frequency utilization, inefficient fusion, and…

arXiv cs.CV 04:00 UTC

arXiv:2610.34917v1 Announce Type: new Abstract: Diffusion models enable multiple advertisers to compete for exposure within a shared generated image. Yet existing approaches are computationally costly when generating images under varying…

arXiv cs.CV 04:00 UTC

arXiv:2603.29441v3 Announce Type: replace Abstract: While the Earth observation community has witnessed a surge in high-impact foundation models and global Earth embedding datasets, a significant barrier remains in translating these…

arXiv cs.CV 04:00 UTC

arXiv:2607.10495v2 Announce Type: replace Abstract: Recent Video Super-Resolution (VSR) methods rely heavily on transformers and explicit optical flow, creating computational overhead and custom operations that hinder deployment on…

arXiv cs.CV 04:00 UTC

arXiv:2606.06498v2 Announce Type: replace-cross Abstract: Traditional statistical graphics are precise but often lack the visual appeal, memorability, and engagement of pictorial charts. We present a generative framework for the automated…

arXiv cs.CL 04:00 UTC

arXiv:2610.34875v1 Announce Type: new Abstract: Rubric-based evaluation asks which level of a rubric a response meets. It is usually done by prompting a large language model (LLM) as a judge, which needs a generative call per judgment and…

arXiv cs.CV 04:00 UTC

arXiv:2606.08258v2 Announce Type: replace-cross Abstract: Understanding and comparing structures in scalar fields is a central challenge in scientific visualization, with applications ranging from feature analysis to temporal and…

arXiv cs.CL 04:00 UTC

arXiv:2610.34212v1 Announce Type: new Abstract: Large language models are increasingly being used in different stages of scientific knowledge production, including the suggestion of references. These models can operate at different levels…

arXiv cs.CL 04:00 UTC

arXiv:2610.35254v1 Announce Type: new Abstract: The number of paper submissions to major AI conferences has increased dramatically in recent years, substantially increasing the peer-review workload of researchers. At the same time, large…

arXiv cs.CV 04:00 UTC

arXiv:2610.35005v1 Announce Type: new Abstract: Intelligent operating rooms require accurate awareness of surgical workflow for better intra-operative decisions, teamwork, and surgical documentation. Therefore, accurately recognizing…

arXiv cs.CL 04:00 UTC

arXiv:2604.28075v3 Announce Type: replace Abstract: Recent research has shown that filtering massive English web corpora into high-quality subsets significantly improves training efficiency. However, for high-resource non-English languages…

arXiv cs.CL 04:00 UTC

arXiv:2606.15910v3 Announce Type: replace Abstract: An answer from a vision-language model (VLM) to a medical image question can be accepted without review only under a threshold bounding the error of accepted answers, and certifying that…

arXiv cs.CV 04:00 UTC

arXiv:2610.35637v1 Announce Type: new Abstract: Due to the inherent discrepancies in imaging viewpoints and field-of-view mismatches among multi-modal sensors, existing fusion methods have long struggled to achieve complete scene-level…

arXiv cs.CL 04:00 UTC

arXiv:2610.33440v1 Announce Type: new Abstract: Selecting a compact training subset is an important step in adapting medical language models from existing question-answering banks. We investigate perturbation-based data selection using two…

arXiv cs.CV 04:00 UTC

arXiv:2605.12567v3 Announce Type: replace Abstract: The inherent electronic noise, side lobes, and speckle noise complicate clinical interpretation of ultrasound images. Conventional denoising methods rely on explicit noise assumptions…

arXiv cs.CV 04:00 UTC

arXiv:2607.00687v2 Announce Type: replace Abstract: Comparing transformer backbones for image segmentation is confounded: each is paired with a different decoder, recipe, and pretraining, so reported differences rarely reflect the backbone…

arXiv cs.CV 04:00 UTC

arXiv:2610.34180v1 Announce Type: cross Abstract: Long-sequence models must retain earlier information. Full attention keeps all past tokens but grows expensive, and recurrent-memory models compress history into fixed slots for linear-time…

arXiv cs.CV 04:00 UTC

arXiv:2610.34795v1 Announce Type: new Abstract: Cities must coordinate domains like infrastructure, mobility, environmental management, emergency response, and public services as conditions evolve. An urban Spatio-Temporal Intelligence…

arXiv cs.CV 04:00 UTC

arXiv:2610.35219v1 Announce Type: new Abstract: Existing multi-agent collaboration (MAC) methods have primarily explored a group thinking collaboration topology, where agents independently generate solutions and aggregate their outputs. In…

arXiv cs.CV 04:00 UTC

arXiv:2610.33038v1 Announce Type: cross Abstract: Chest computed tomography (CT) reports generated by large language models (LLMs) can describe positive findings that lack localized evidence in the scan. We propose EviGround-CT to verify…

arXiv cs.CL 04:00 UTC

arXiv:2504.11952v4 Announce Type: replace Abstract: An ideal detection system for machine generated content is supposed to work well on any generator as many more advanced LLMs come into existence day by day. Existing systems often…

arXiv cs.CL 04:00 UTC

arXiv:2610.34537v1 Announce Type: new Abstract: Estimating the correctness of large language model (LLM) predictions without access to ground truth is critical for reliable deployment in high-stakes settings. Existing approaches, such as…

arXiv cs.CL 04:00 UTC

arXiv:2606.05405v4 Announce Type: replace-cross Abstract: Recent AI systems have achieved strong results on a wide range of benchmarks, yet these gains have not translated into economically meaningful deployment across many professional…

arXiv cs.CL 04:00 UTC

arXiv:2610.33765v1 Announce Type: new Abstract: Streaming automatic speech recognition transcribes at low latency but produces an unattributed word stream, while real-time applications such as customer-service analytics need each word tied…

arXiv cs.CV 04:00 UTC

arXiv:2606.13315v2 Announce Type: replace Abstract: MRI self-supervised learning (SSL) has largely emphasized segmentation and dense prediction, while controlled studies of how alternative pretraining objectives shape transferable…

arXiv cs.CV 04:00 UTC

arXiv:2605.01165v2 Announce Type: replace Abstract: This paper proposes a novel Zero-Shot Action Recognition~(ZSAR) method based on contrastive learning. In ZSAR, we aim to classify examples from classes that were missing during training…

arXiv cs.CL 04:00 UTC

arXiv:2606.02423v2 Announce Type: replace Abstract: Large language models (LLMs) can serve as helpful assistants, yet they can equally function as harm amplifiers that enable malicious users to achieve harmful outcomes beyond their…

arXiv cs.CV 04:00 UTC

arXiv:2610.35501v1 Announce Type: new Abstract: Learning-based structure-from-motion methods such as ACE-Zero have demonstrated strong performance in estimating cameraposes and scene coordinates from unordered image collections without…

arXiv cs.CV 04:00 UTC

arXiv:2610.35450v1 Announce Type: new Abstract: Referring expression grounding in overhead imagery requires oriented bounding boxes rather than horizontal ones, since aerial objects appear at arbitrary orientations. Most grounding models…

arXiv cs.CV 04:00 UTC

arXiv:2610.35515v1 Announce Type: new Abstract: Diffusion-based real-world image super-resolution suffers from limited reconstruction fidelity due to information loss in VAE-based latent representations. Recently, pixel-space diffusion…

arXiv cs.CV 04:00 UTC

arXiv:2610.34033v1 Announce Type: new Abstract: Camera motion recognition requires separating viewpoint changes caused by camera translation, rotation, or zoom from object motion and scene dynamics. Video-language models often rely on…

arXiv cs.CV 04:00 UTC

arXiv:2610.34163v1 Announce Type: new Abstract: Crowd barriers are widely used to regulate pedestrian movement during large public events such as pilgrimages. However, fixed barrier layouts often create unnecessary stopping and local…

arXiv cs.CV 04:00 UTC

arXiv:2610.35766v1 Announce Type: cross Abstract: Self-supervised training is monitored without labels, through the loss and statistics of the learned features. We report a case where we know the encoders lost part of what they had…

arXiv cs.CL 04:00 UTC

arXiv:2610.33016v1 Announce Type: new Abstract: Llama-3, Qwen3 and many other open LLMs pre-tokenize text with a regular expression whose word class admits only Unicode letters. Devanagari vowel signs are combining marks, so this regex…

arXiv cs.CL 04:00 UTC

arXiv:2610.33512v1 Announce Type: new Abstract: Rubric-based rewards enable effective reinforcement learning on non-verifiable tasks. However, because these rewards are typically defined as a weighted measure of rubric coverage, models are…

arXiv cs.CL 04:00 UTC

arXiv:2610.33608v1 Announce Type: new Abstract: Hallucination detection is increasingly important for evaluating and deploying large language models. However, existing methods are often compared across benchmarks that define and measure…

arXiv cs.CL 04:00 UTC

arXiv:2610.35846v1 Announce Type: new Abstract: Expanding multilingual ASR systems to new languages is challenging when adaptation must preserve previously supported languages without replaying source-language training data. We study this…

arXiv cs.CL 04:00 UTC

arXiv:2610.33904v1 Announce Type: cross Abstract: Large language models (LLMs) can exploit shortcuts, i.e., correlations present in training data that may not hold at test time. This can lead to poor predictions when correlations shift…

arXiv cs.CV 04:00 UTC

arXiv:2601.19128v3 Announce Type: replace Abstract: In terrestrial laser scanning (TLS)-based mechanical, electrical, and plumbing (MEP) point cloud segmentation, safety-critical components such as reducers and valves are persistently…

arXiv cs.CL 04:00 UTC

arXiv:2605.23420v2 Announce Type: replace Abstract: Social norms reflect shared expectations on acceptable behavior. Measuring social norms alignment remains challenging, with existing approaches typically relying on artificial closed-form…

arXiv cs.CL 04:00 UTC

arXiv:2604.00004v3 Announce Type: replace Abstract: The extension of context windows in Large Language Models is typically facilitated by scaling positional encodings followed by Continued Pre-Training (CPT). While effective, this paradigm…

arXiv cs.CL 04:00 UTC

arXiv:2607.07408v3 Announce Type: replace Abstract: Automatic prosodic segmentation identifies boundaries between speech units based on acoustic and linguistic evidence. Although recent deep learning approaches have achieved strong results…

arXiv cs.CL 04:00 UTC

arXiv:2610.33013v1 Announce Type: cross Abstract: Multi-Agent Systems (MAS) are a promising paradigm for improving the reasoning capabilities of large language models, yet they remain prone to hallucinations and erroneous outputs…

arXiv cs.CL 04:00 UTC

arXiv:2610.34457v1 Announce Type: cross Abstract: Speech and language technologies have expanded access to communication, but the support they provide is often fixed. The same linguistic or visual aid may help one user and hinder another…

arXiv cs.CL 04:00 UTC

arXiv:2610.34491v1 Announce Type: cross Abstract: Generative models increasingly expose a reasoning effort setting that lets users scale inference-time compute, trading higher cost for potentially improved performance. Users must choose an…

arXiv cs.CL 04:00 UTC

arXiv:2607.03863v3 Announce Type: replace Abstract: Artificial intelligence has advanced scientific discovery, but most AI4Science systems remain fragmented tools that rely on humans to coordinate problem formulation, literature grounding…

arXiv cs.CL 04:00 UTC

arXiv:2610.33813v1 Announce Type: new Abstract: Fine-tuning a language model for classification involves many design decisions. Comparisons of low-rank adaptation (LoRA) with full fine-tuning are common but often confounded, and task…

arXiv cs.CL 04:00 UTC

arXiv:2610.34746v1 Announce Type: new Abstract: In high-stakes domains such as finance, critical information evolves through multi-party interaction. Recovering these changes and their sources is fundamental to reliable speech…

arXiv cs.CL 04:00 UTC

arXiv:2610.35029v1 Announce Type: new Abstract: Synthetic data has become a common resource for training language models, and this trend has spilled over into research on developmentally plausible language modeling. However, it remains…

arXiv cs.CV 04:00 UTC

arXiv:2610.34926v1 Announce Type: new Abstract: An action seen once and the same action seen three times receive the same action-recognition label, so that supervision does not require a model to count. Answers shared across histories that…

arXiv cs.CL 04:00 UTC

arXiv:2610.34991v1 Announce Type: new Abstract: End-to-end full-duplex spoken dialogue models enable natural, real-time spoken interactions with interruptions, backchannels, and overlapping speech. Prior evaluations have assessed diverse…

arXiv cs.CL 04:00 UTC

arXiv:2610.33766v1 Announce Type: new Abstract: Users of conversational assistants rarely state all requirements up front: they add constraints, amend them, withdraw them, digress, and return. Instruction-tuned LLMs fail here in two ways…

arXiv cs.CL 04:00 UTC

arXiv:2610.33749v1 Announce Type: new Abstract: As machine translation systems approach human-level quality, traditional evaluation methodologies struggle to detect subtle translation errors. We critically examine limitations in current…

arXiv cs.CL 04:00 UTC

arXiv:2610.34768v1 Announce Type: new Abstract: Knowledge editing aims to update outdated or incorrect knowledge in large language models without full retraining. Existing parameter-editing methods remain limited for long-sequence…

arXiv cs.CL 04:00 UTC

arXiv:2610.33001v1 Announce Type: new Abstract: Academic revision presents a governance problem when legitimate writing objectives conflict. Writing-process and social-science research show that revision is recursive and that wording…

arXiv cs.CL 04:00 UTC

arXiv:2610.35685v1 Announce Type: new Abstract: Large language model (LLM) pipelines can turn visitor reviews into structured data, but generic schemes may miss rituals, sacred spaces, cultural restrictions and emotionally mixed…

arXiv cs.CL 04:00 UTC

arXiv:2610.35697v1 Announce Type: cross Abstract: Error-targeted learning gives a legible way to improve a model: identify specific weaknesses of the model, train the model to fix these weaknesses, and check whether the weaknesses are…

arXiv cs.CL 04:00 UTC

arXiv:2607.16051v3 Announce Type: replace Abstract: We present Loopie, the most powerful open-source looped Transformer to date. The Loopie series consists of two Mixture-of-Experts (MoE) models: a 20B-parameter model with 2B active…

arXiv cs.CL 04:00 UTC

arXiv:2610.35804v1 Announce Type: new Abstract: Diffusion large language models (dLLMs) can generate multiple tokens in parallel, but often require many decoding rounds to complete a response. Efficient generation depends not only on which…

arXiv cs.CV 04:00 UTC

arXiv:2610.33641v1 Announce Type: new Abstract: Attribute-based person retrieval ranks surveillance images from a binary query of semantic attributes, but mean Average Attribute Degree of Match (mADM) drops on unseen cameras as predicted…

arXiv cs.CV 04:00 UTC

arXiv:2610.35152v1 Announce Type: new Abstract: Recent studies on generalizable human Gaussian splatting have achieved the notable progress in synthesizing novel views of unseen human subjects from multi-view images. Despite this progress…

arXiv cs.CV 04:00 UTC

arXiv:2601.11990v2 Announce Type: replace Abstract: Multimodal action understanding requires complementary sensory observations to be connected with structured semantic knowledge. This is particularly challenging in vehicle cabins, where…

arXiv cs.CV 04:00 UTC

arXiv:2604.10383v4 Announce Type: replace Abstract: Authoring a multi-actor scenario for a living 3D world, where every action changes its state, and each action's validity depends on the state accumulated before it, demands the freedom of…

arXiv cs.CV 04:00 UTC

arXiv:2610.35536v1 Announce Type: new Abstract: Uniform frame sampling can miss brief evidence in sparse-event videos when the VLM sees only K frames. We propose EARS (Evidence-Aware Retrieval via Sound and motion), a training-free…

arXiv cs.CV 04:00 UTC

arXiv:2610.33078v1 Announce Type: new Abstract: Visual grounding (VG) in visual question answering (VQA) is a fundamental capability of Vision-Language Models (VLMs), enabling the detection of the most relevant region in an image to answer…

arXiv cs.CV 04:00 UTC

arXiv:2610.33941v1 Announce Type: new Abstract: World action models have demonstrated promising performance in robot policy learning, yet their high inference latency remains a critical bottleneck for real-world deployment. Recent…

arXiv cs.CV 04:00 UTC

arXiv:2610.34214v1 Announce Type: new Abstract: Cross-modality medical image translation is challenged by anatomically inconsistent synthesis that can compromise clinical diagnosis. We propose StaX-Net, a Stackelberg interleaved asymmetric…

arXiv cs.CV 04:00 UTC

arXiv:2610.34766v1 Announce Type: new Abstract: Diffeomorphisms provide a flexible, topology-preserving mathematical tool for modeling complex spatial or temporal deformations and are used in various scientific fields. However, making…

arXiv cs.CV 04:00 UTC

arXiv:2610.35396v1 Announce Type: new Abstract: We evaluated synthetic pretraining and multi-cohort transfer learning for deep learning-based super-resolution of intracranial 4D Flow MRI across heterogeneous neurologic cohorts. This mixed…

arXiv cs.CV 04:00 UTC

arXiv:2610.35482v1 Announce Type: cross Abstract: Social media enables cross-lingual communication, yet the meaning of user-generated content (UGC) is often distributed across text, images, and cultural knowledge, making translation…

arXiv cs.CV 04:00 UTC

arXiv:2511.20251v3 Announce Type: replace Abstract: Modern text-to-image models produce impressive visual results from richly specified prompts, yet their behavior under long prompts remains insufficiently understood. In this paper, we…

arXiv cs.CV 04:00 UTC

arXiv:2610.34785v1 Announce Type: new Abstract: Deep neural networks are known to rely on non-robust features that align poorly with human perception. Prior work has largely framed these features as opportunistic shortcuts that help…

arXiv cs.CV 04:00 UTC

arXiv:2605.28036v2 Announce Type: replace Abstract: Diffusion models steer conditional generation with a tunable guidance scale to trade off prompt alignment and diversity. However, existing debiasing techniques are optimized for a single…

arXiv cs.CL 04:00 UTC

arXiv:2610.35556v1 Announce Type: new Abstract: Looped Transformers increase computation per token by repeatedly applying a shared backbone, offering a way to improve language modeling and downstream task performance at a fixed backbone…

arXiv cs.CL 04:00 UTC

arXiv:2602.10657v2 Announce Type: replace Abstract: How much of zero-shot benchmark accuracy is captured by word-level alignment between a pre-training corpus and the benchmark target? We study this question through word-level unigram…

arXiv cs.CV 04:00 UTC

arXiv:2607.18451v4 Announce Type: replace-cross Abstract: In medical imaging, one frozen foundation encoder is often reused for many tasks. The encoder is usually chosen by the area under the receiver operating characteristic curve (AUROC)…

arXiv cs.CV 04:00 UTC

arXiv:2610.33045v1 Announce Type: new Abstract: Integration of artificial intelligence (AI) technology into identity workflows, such as automated face recognition (FR), has enhanced operational capabilities, with FR algorithms achieving…

arXiv cs.CV 04:00 UTC

arXiv:2610.33635v1 Announce Type: new Abstract: Pavement crack segmentation under low light is degraded by reduced contrast, shadows, headlight glare, joints, stains, and wet seams. Existing crack segmentation methods mostly assume good…

arXiv cs.CV 04:00 UTC

arXiv:2610.33894v1 Announce Type: new Abstract: Interactive world models hold immense potential for high-stakes applications in robotics, game engines, and immersive digital twins, yet existing benchmarks primarily focus on perceptual…

arXiv cs.CV 04:00 UTC

arXiv:2610.34321v1 Announce Type: new Abstract: Text-to-motion models generate for the skeletons they were trained on, while a new character arrives as a skeleton file and nothing else. We study text-to-motion for skeletons that no…

arXiv cs.CV 04:00 UTC

arXiv:2610.33124v1 Announce Type: cross Abstract: 4D Flow MRI promises opportunities for non-invasive mapping of neurovascular hemodynamics, but its clinical utility faces trade-offs between resolution, acquisition time, and noise. Deep…

arXiv cs.CV 04:00 UTC

arXiv:2610.33430v1 Announce Type: new Abstract: Recent DETR-based tiny object detectors adopt dynamic query mechanisms to handle severe density imbalance. However, existing designs entangle two effects: how many and which queries are…

arXiv cs.CV 04:00 UTC

arXiv:2610.34028v1 Announce Type: new Abstract: As Vision-Language Models (VLMs) increasingly enable robots to follow language instructions, understanding their navigation decisions becomes important for reliable deployment. Existing…

arXiv cs.CV 04:00 UTC

arXiv:2603.18558v3 Announce Type: replace Abstract: Long-form video question answering requires reasoning over extended temporal contexts, making frame selection a critical bottleneck for multi-modal large language models (MLLMs) bound by…

arXiv cs.CV 04:00 UTC

arXiv:2606.13894v4 Announce Type: replace-cross Abstract: AdamW is a default optimizer for deep learning, but its moment states add two parameter-sized buffers to training memory, increasing the cost of large-scale pretraining. We propose…

arXiv cs.CV 04:00 UTC

arXiv:2610.33484v1 Announce Type: new Abstract: Text-to-3D solid retrieval requires aligning natural language with precise geometry despite variations in pose, scale, and recording conventions. Existing methods discretize Boundary…

arXiv cs.CV 04:00 UTC

arXiv:2610.33580v1 Announce Type: new Abstract: Serial electron microscopy (EM) is essential for large-volume biological imaging and connectomics, yet its throughput remains a major bottleneck, with cubic-millimeter-scale brain imaging…

arXiv cs.CV 04:00 UTC

arXiv:2610.33966v1 Announce Type: cross Abstract: People have long dreamed of an assistant that is always there to guide them and remind what they forget, from new dishes to lost keys. Powered by multimodal large language models (MLLMs)…

arXiv cs.CL 04:00 UTC

arXiv:2601.06580v3 Announce Type: replace Abstract: Singlish is a creole rooted in Singapore's multilingual environment that continues to evolve alongside social and technological change. We examine diachronic stylistic change across a…

arXiv cs.CV 04:00 UTC

arXiv:2607.07038v2 Announce Type: replace Abstract: Glioma segmentation models can lose accuracy across MRI datasets, while identifying the cases that require correction remains difficult. In this study, we introduce counterfactual context…

arXiv cs.CL 04:00 UTC

arXiv:2610.33488v1 Announce Type: new Abstract: We evaluate whether semester-level interaction features from a large language model tutor improve assignment-score prediction beyond constant baselines. StudyChat provides 16,851…

arXiv cs.CL 04:00 UTC

arXiv:2610.33583v1 Announce Type: new Abstract: Multilingual BabyLM systems usually fix language sampling proportions even though the weakest language can change during training. We present ARMOR-DP, a modular framework that diagnoses…

arXiv cs.CL 04:00 UTC

arXiv:2610.33586v1 Announce Type: new Abstract: Depression severity estimation is crucial for mental health assessment and intervention, yet traditional approaches depend on clinical interviews or self-report questionnaires that require…

arXiv cs.CL 04:00 UTC

arXiv:2610.33806v1 Announce Type: new Abstract: Language-model agents must adapt as users revise their requirements and priorities during complex tasks. We introduce IntentDelta, a human-constructed multi-turn benchmark with 638…

arXiv cs.CL 04:00 UTC

arXiv:2610.33962v1 Announce Type: new Abstract: Tool-using large language models (LLMs) must decide whether external information warrants action. We investigate whether their expressed credibility judgments reliably indicate behavioral…

arXiv cs.CL 04:00 UTC

arXiv:2610.34653v1 Announce Type: new Abstract: Latency-sensitive spoken agents must know not only which tool to call, but also when they have heard enough to act. Waiting for the turn to complete delays execution even when the request is…

arXiv cs.CL 04:00 UTC

arXiv:2610.35102v1 Announce Type: new Abstract: Synthetic supervision can bootstrap semantic parsers in new search domains, but prompt-induced diversity is useful only when generations are faithful to the structured intent. We explore this…

arXiv cs.CL 04:00 UTC

arXiv:2610.35201v1 Announce Type: new Abstract: Diffusion language models (DLMs) generate text through iterative denoising, enabling bidirectional context modeling and parallel token refinement. Recent studies have applied reinforcement…