OpenAI is close to finalizing a 40 billion ( 6 trillion) funding round led by SoftBank Group -- with investors including Magnetar Capital, Coatue Management, Founders Fund and Altimeter Capital Management in talks to participate, according to people familiar with the matter. Magnetar Capital -- an Evanston, Illinois-based hedge fund -- could contribute up to 1 billion, according to multiple people, all of whom asked not to be identified because the information is private. The artificial intelligence developer's funding round would be the largest of all time, according to data compiled by research firm PitchBook. The deal is set to value the company at 300 billion including dollars raised -- almost double the ChatGPT maker's previous valuation of 157 billion from when it raised money in October.

Visual in-context learning (VICL), as a new paradigm in computer vision, allows the model to rapidly adapt to various tasks with only a handful of prompts and examples. While effective, the existing VICL paradigm exhibits poor generalizability under distribution shifts. In this work, we propose test-time Visual In-Context Tuning (VICT), a method that can adapt VICL models on the fly with a single test sample. Specifically, we flip the role between the task prompts and the test sample and use a cycle consistency loss to reconstruct the original task prompt output. Our key insight is that a model should be aware of a new test distribution if it can successfully recover the original task prompts. Extensive experiments on six representative vision tasks ranging from high-level visual understanding to low-level image processing, with 15 common corruptions, demonstrate that our VICT can improve the generalizability of VICL to unseen new domains. In addition, we show the potential of applying VICT for unseen tasks at test time. Code: https://github.com/Jiahao000/VICT.

Intent, typically clearly formulated and planned, functions as a cognitive framework for reasoning and problem-solving. This paper introduces the concept of Speaking with Intent (SWI) in large language models (LLMs), where the explicitly generated intent encapsulates the model's underlying intention and provides high-level planning to guide subsequent analysis and communication. By emulating deliberate and purposeful thoughts in the human mind, SWI is hypothesized to enhance the reasoning capabilities and generation quality of LLMs. Extensive experiments on mathematical reasoning benchmarks consistently demonstrate the superiority of Speaking with Intent over Baseline (i.e., generation without explicit intent). Moreover, SWI outperforms answer-trigger prompting methods Chain-of-Thought and Plan-and-Solve and maintains competitive performance with the strong method ARR (Analyzing, Retrieving, and Reasoning). Additionally, the effectiveness and generalizability of SWI are solidified on reasoning-intensive question answering (QA) and text summarization benchmarks, where SWI brings consistent improvement to the Baseline generation. In text summarization, SWI-generated summaries exhibit greater accuracy, conciseness, and factual correctness, with fewer hallucinations. Furthermore, human evaluations verify the coherence, effectiveness, and interpretability of the intent produced by SWI. This proof-of-concept study creates a novel avenue for enhancing LLMs' reasoning abilities with cognitive notions.

Abstract--While large language models (LLMs) are powerful assistants in programming tasks, they may also produce mali cious code. T esting LLM-generated code therefore poses significa nt risks to assessment infrastructure tasked with executing u n-trusted code. T o address these risks, this work focuses on evaluating the security and confidentiality properties of t est environments, reducing the risk that LLM-generated code ma y compromise the assessment infrastructure. We introduce Sa nd-boxEval, a test suite featuring manually crafted test cases that simulate real-world safety scenarios for LLM assessment en vi-ronments in the context of untrusted code execution. The sui te evaluates vulnerabilities to sensitive information expos ure, filesys-tem manipulation, external communication, and other poten tially dangerous operations in the course of assessment activity. We demonstrate the utility of SandboxEval by deploying it on an open-source implementation of Dyff, an established AI asse ssment framework used to evaluate the safety of LLMs at scale. We sho w, first, that the test suite accurately describes limitations placed on an LLM operating under instructions to generate malicious c ode. Second, we show that the test results provide valuable insig hts for developers seeking to harden assessment infrastructur e and identify risks associated with LLM execution activities. There is growing interest in using large language models (LLMs) to assist with code generation due to their ability to produce relevant code for various programming tasks [1, 2, 3 ]. However, using code generated by LLMs involves certain risks, as it may contain subtle bugs or security flaws that are not immediately apparent [4, 5, 6, 7]. For instance, a malicious model developer may intentionally train poisone d LLMs to inject malicious code, subtly manipulating completions to benefit themselves [8, 9, 10]. Through prompt injection, a malicious user may manipulate input to produce harmful outputs [11, 12].

Language models (LMs) have become a staple of the code-writing toolbox. Their pre-training recipe has, however, remained stagnant over recent years, barring the occasional changes in data sourcing and filtering strategies. In particular, research exploring modifications to Code-LMs' pre-training objectives, geared towards improving data efficiency and better disentangling between syntax and semantics, has been noticeably sparse, especially compared with corresponding efforts in natural language LMs. In this work, we examine grounding on obfuscated code as a means of helping Code-LMs look beyond the surface-form syntax and enhance their pre-training sample efficiency. To this end, we compile ObscuraX, a dataset of approximately 55M source and obfuscated code pairs in seven languages. Subsequently, we pre-train ObscuraCoder models, ranging in size from 255M to 2.8B parameters, on a 272B-token corpus that includes ObscuraX and demonstrate that our obfuscation-based pre-training recipe leads to consistent improvements in Code-LMs' abilities compared to both vanilla autoregressive pre-training as well as existing de-obfuscation (DOBF) objectives. ObscuraCoder demonstrates sizeable gains across multiple tests of syntactic and semantic code understanding, along with improved capabilities in multilingual code completion, multilingual code commit summarization, and multi-purpose library-oriented code generation.

In recent years, large language models (LLMs) have shown significant advancements in natural language processing (NLP), with strong capabilities in generation, comprehension, and reasoning. These models have found applications in education, intelligent decision-making, and gaming. However, effectively utilizing LLMs for strategic planning and decision-making in the game of Gomoku remains a challenge. This study aims to develop a Gomoku AI system based on LLMs, simulating the human learning process of playing chess. The system is designed to understand and apply Gomoku strategies and logic to make rational decisions. The research methods include enabling the model to "read the board," "understand the rules," "select strategies," and "evaluate positions," whil e enhancing its abilities through self -play and reinforcement learning. The results demonstrate that this approach significantly improves the selection of move positions, resolves the issue of generating illegal positions, and reduces process time through parallel position evaluation. After extensive self -play training, the model's Gomoku-playing capabilities have been notably enhanced.

Integrating large language models (LLMs) like DeepSeek R1 into healthcare requires rigorous evaluation of their reasoning alignment with clinical expertise. This study assesses DeepSeek R1's medical reasoning against expert patterns using 100 MedQA clinical cases. The model achieved 93% diagnostic accuracy, demonstrating systematic clinical judgment through differential diagnosis, guideline-based treatment selection, and integration of patient-specific factors. However, error analysis of seven incorrect cases revealed persistent limitations: anchoring bias, challenges reconciling conflicting data, insufficient exploration of alternatives, overthinking, knowledge gaps, and premature prioritization of definitive treatment over intermediate care. Crucially, reasoning length correlated with accuracy - shorter responses (<5,000 characters) were more reliable, suggesting extended explanations may signal uncertainty or rationalization of errors. While DeepSeek R1 exhibits foundational clinical reasoning capabilities, recurring flaws highlight critical areas for refinement, including bias mitigation, knowledge updates, and structured reasoning frameworks. These findings underscore LLMs' potential to augment medical decision-making through artificial reasoning but emphasize the need for domain-specific validation, interpretability safeguards, and confidence metrics (e.g., response length thresholds) to ensure reliability in real-world applications.

Composed Image Retrieval (CIR) facilitates image retrieval through a multimodal query consisting of a reference image and modification text. The reference image defines the retrieval context, while the modification text specifies desired alterations. However, existing CIR datasets predominantly employ coarse-grained modification text (CoarseMT), which inadequately captures fine-grained retrieval intents. This limitation introduces two key challenges: (1) ignoring detailed differences leads to imprecise positive samples, and (2) greater ambiguity arises when retrieving visually similar images. These issues degrade retrieval accuracy, necessitating manual result filtering or repeated queries. To address these limitations, we develop a robust fine-grained CIR data annotation pipeline that minimizes imprecise positive samples and enhances CIR systems' ability to discern modification intents accurately. Using this pipeline, we refine the FashionIQ and CIRR datasets to create two fine-grained CIR datasets: Fine-FashionIQ and Fine-CIRR. Furthermore, we introduce FineCIR, the first CIR framework explicitly designed to parse the modification text. FineCIR effectively captures fine-grained modification semantics and aligns them with ambiguous visual entities, enhancing retrieval precision. Extensive experiments demonstrate that FineCIR consistently outperforms state-of-the-art CIR baselines on both fine-grained and traditional CIR benchmark datasets. Our FineCIR code and fine-grained CIR datasets are available at https://github.com/SDU-L/FineCIR.git.

Although applying Mixture of Experts to large language models for learning new tasks is widely regarded as an effective strategy for continuous learning, there still remain two major challenges: (1) As the number of tasks grows, simple parameter expansion strategies can lead to excessively large models. (2) Modifying the parameters of the existing router results in the erosion of previously acquired knowledge. In this paper, we present an innovative framework named LLaVA-CMoE, which is a continuous Mixture of Experts (MoE) architecture without any replay data. Specifically, we have developed a method called Probe-Guided Knowledge Extension (PGKE), which employs probe experts to assess whether additional knowledge is required for a specific layer. This approach enables the model to adaptively expand its network parameters based on task distribution, thereby significantly improving the efficiency of parameter expansion. Additionally, we introduce a hierarchical routing algorithm called Probabilistic Task Locator (PTL), where high-level routing captures inter-task information and low-level routing focuses on intra-task details, ensuring that new task experts do not interfere with existing ones. Our experiments shows that our efficient architecture has substantially improved model performance on the Coin benchmark while maintaining a reasonable parameter count.

Large language models have recently transformed hardware design, yet bridging the gap between code synthesis and PPA (power, performance, and area) estimation remains a challenge. In this work, we introduce a novel framework that leverages a 21k dataset of thoroughly cleaned and synthesizable Verilog modules, each annotated with detailed power, delay, and area metrics. By employing chain-of-thought techniques, we automatically debug and curate this dataset to ensure high fidelity in downstream applications. We then fine-tune CodeLlama using LoRA-based parameter-efficient methods, framing the task as a regression problem to accurately predict PPA metrics from Verilog code. Furthermore, we augment our approach with a mixture-of-experts architecture-integrating both LoRA and an additional MLP expert layer-to further refine predictions. Experimental results demonstrate significant improvements: power estimation accuracy is enhanced by 5.9% at a 20% error threshold and by 7.2% at a 10% threshold, delay estimation improves by 5.1% and 3.9%, and area estimation sees gains of 4% and 7.9% for the 20% and 10% thresholds, respectively. Notably, the incorporation of the mixture-of-experts module contributes an additional 3--4% improvement across these tasks. Our results establish a new benchmark for PPA-aware Verilog generation, highlighting the effectiveness of our integrated dataset and modeling strategies for next-generation EDA workflows.