Ambiguity is pervasive in real-world questions, yet large language models (LLMs) often respond with confident answers rather than seeking clarification. In this work, we show that question ambiguity is linearly encoded in the internal representations of LLMs and can be both detected and controlled at the neuron level. During the model's pre-filling stage, we identify that a small number of neurons, as few as one, encode question ambiguity information. Probes trained on these Ambiguity-Encoding Neurons (AENs) achieve strong performance on ambiguity detection and generalize across datasets, outperforming prompting-based and representation-based baselines. Layerwise analysis reveals that AENs emerge from shallow layers, suggesting early encoding of ambiguity signals in the model's processing pipeline. Finally, we show that through manipulating AENs, we can control LLM's behavior from direct answering to abstention. Our findings reveal that LLMs form compact internal representations of question ambiguity, enabling interpretable and controllable behavior.

As software development practices increasingly adopt AI-powered tools, ensuring that such tools can support secure coding has become critical. This study evaluates the effectiveness of GitHub Copilot's recently introduced code review feature in detecting security vulnerabilities. Using a curated set of labeled vulnerable code samples drawn from diverse open-source projects spanning multiple programming languages and application domains, we systematically assessed Copilot's ability to identify and provide feedback on common security flaws. Contrary to expectations, our results reveal that Copilot's code review frequently fails to detect critical vulnerabilities such as SQL injection, cross-site scripting (XSS), and insecure deserialization. Instead, its feedback primarily addresses low-severity issues, such as coding style and typographical errors. These findings expose a significant gap between the perceived capabilities of AI-assisted code review and its actual effectiveness in supporting secure development practices. Our results highlight the continued necessity of dedicated security tools and manual code audits to ensure robust software security.

As Large Multimodal Models (LMMs) become integral to daily digital life, understanding their safety architectures is a critical problem for AI Alignment. This paper presents a systematic analysis of OpenAI's GPT-4o mini, a globally deployed model, on the difficult task of multimodal hate speech detection. Using the Hateful Memes Challenge dataset, we conduct a multi-phase investigation on 500 samples to probe the model's reasoning and failure modes. Our central finding is the experimental identification of a "Unimodal Bottleneck," an architectural flaw where the model's advanced multimodal reasoning is systematically preempted by context-blind safety filters. A quantitative validation of 144 content policy refusals reveals that these overrides are triggered in equal measure by unimodal visual 50% and textual 50% content. We further demonstrate that this safety system is brittle, blocking not only high-risk imagery but also benign, common meme formats, leading to predictable false positives. These findings expose a fundamental tension between capability and safety in state-of-the-art LMMs, highlighting the need for more integrated, context-aware alignment strategies to ensure AI systems can be deployed both safely and effectively.

Beyond general web-scale search, social network search uniquely enables users to retrieve information and discover potential connections within their social context. We introduce a framework of modernized Facebook Group Scoped Search by blending traditional keyword-based retrieval with embedding-based retrieval (EBR) to improve the search relevance and diversity of search results. Our system integrates semantic retrieval into the existing keyword search pipeline, enabling users to discover more contextually relevant group posts. To rigorously assess the impact of this blended approach, we introduce a novel evaluation framework that leverages large language models (LLMs) to perform offline relevance assessments, providing scalable and consistent quality benchmarks. Our results demonstrate that the blended retrieval system significantly enhances user engagement and search quality, as validated by both online metrics and LLM-based evaluation. This work offers practical insights for deploying and evaluating advanced retrieval systems in large-scale, real-world social platforms.

This study examines the potential of utilizing Vision Language Models (VLMs) to improve the perceptual capabilities of semi-autonomous prosthetic hands. We introduce a unified benchmark for end-to-end perception and grasp inference, evaluating a single VLM to perform tasks that traditionally require complex pipelines with separate modules for object detection, pose estimation, and grasp planning. To establish the feasibility and current limitations of this approach, we benchmark eight contemporary VLMs on their ability to perform a unified task essential for bionic grasping. From a single static image, they should (1) identify common objects and their key properties (name, shape, orientation, and dimensions), and (2) infer appropriate grasp parameters (grasp type, wrist rotation, hand aperture, and number of fingers). A corresponding prompt requesting a structured JSON output was employed with a dataset of 34 snapshots of common objects. Key performance metrics, including accuracy for categorical attributes (e.g., object name, shape) and errors in numerical estimates (e.g., dimensions, hand aperture), along with latency and cost, were analyzed. The results demonstrated that most models exhibited high performance in object identification and shape recognition, while accuracy in estimating dimensions and inferring optimal grasp parameters, particularly hand rotation and aperture, varied more significantly. This work highlights the current capabilities and limitations of VLMs as advanced perceptual modules for semi-autonomous control of bionic limbs, demonstrating their potential for effective prosthetic applications.

The rapid advancement of Large Language Models (LLMs) has intensified the need for robust dialogue system evaluation, yet comprehensive assessment remains challenging. Traditional metrics often prove insufficient, and safety considerations are frequently narrowly defined or culturally biased. The DSTC12 Track 1, "Dialog System Evaluation: Dimensionality, Language, Culture and Safety," is part of the ongoing effort to address these critical gaps. The track comprised two subtasks: (1) Dialogue-level, Multi-dimensional Automatic Evaluation Metrics, and (2) Multilingual and Multicultural Safety Detection. For Task 1, focused on 10 dialogue dimensions, a Llama-3-8B baseline achieved the highest average Spearman's correlation (0.1681), indicating substantial room for improvement. In Task 2, while participating teams significantly outperformed a Llama-Guard-3-1B baseline on the multilingual safety subset (top ROC-AUC 0.9648), the baseline proved superior on the cultural subset (0.5126 ROC-AUC), highlighting critical needs in culturally-aware safety. This paper describes the datasets and baselines provided to participants, as well as submission evaluation results for each of the two proposed subtasks.

The U.S. Federal Open Market Committee (FOMC) regularly discusses and sets monetary policy, affecting the borrowing and spending decisions of millions of people. In this work, we release Op-Fed, a dataset of 1044 human-annotated sentences and their contexts from FOMC transcripts. We faced two major technical challenges in dataset creation: imbalanced classes -- we estimate fewer than 8% of sentences express a non-neutral stance towards monetary policy -- and inter-sentence dependence -- 65% of instances require context beyond the sentence-level. To address these challenges, we developed a five-stage hierarchical schema to isolate aspects of opinion, monetary policy, and stance towards monetary policy as well as the level of context needed. Second, we selected instances to annotate using active learning, roughly doubling the number of positive instances across all schema aspects. Using Op-Fed, we found a top-performing, closed-weight LLM achieves 0.80 zero-shot accuracy in opinion classification but only 0.61 zero-shot accuracy classifying stance towards monetary policy -- below our human baseline of 0.89. We expect Op-Fed to be useful for future model training, confidence calibration, and as a seed dataset for future annotation efforts.

Generative machine learning offers new opportunities to better understand complex Earth system dynamics. Recent diffusion-based methods address spectral biases and improve ensemble calibration in weather forecasting compared to deterministic methods, yet have so far proven difficult to scale stably at high resolutions. We introduce AERIS, a 1.3 to 80B parameter pixel-level Swin diffusion transformer to address this gap, and SWiPe, a generalizable technique that composes window parallelism with sequence and pipeline parallelism to shard window-based transformers without added communication cost or increased global batch size. On Aurora (10,080 nodes), AERIS sustains 10.21 ExaFLOPS (mixed precision) and a peak performance of 11.21 ExaFLOPS with $1 \times 1$ patch size on the 0.25° ERA5 dataset, achieving 95.5% weak scaling efficiency, and 81.6% strong scaling efficiency. AERIS outperforms the IFS ENS and remains stable on seasonal scales to 90 days, highlighting the potential of billion-parameter diffusion models for weather and climate prediction.

Gender-neutral rewriting (GNR) aims to reformulate text to eliminate unnecessary gender specifications while preserving meaning, a particularly challenging task in grammatical-gender languages like Italian. In this work, we conduct the first systematic evaluation of state-of-the-art large language models (LLMs) for Italian GNR, introducing a two-dimensional framework that measures both neutrality and semantic fidelity to the input. We compare few-shot prompting across multiple LLMs, fine-tune selected models, and apply targeted cleaning to boost task relevance. Our findings show that open-weight LLMs outperform the only existing model dedicated to GNR in Italian, whereas our fine-tuned models match or exceed the best open-weight LLM's performance at a fraction of its size. Finally, we discuss the trade-off between optimizing the training data for neutrality and meaning preservation.

Large language models (LLMs) show promise for translating natural-language statutes into executable logic, but reliability in legally critical settings remains challenging due to ambiguity and hallucinations. We present an agentic approach for developing legal-critical software, using U.S. federal tax preparation as a case study. The key challenge is test-case generation under the oracle problem, where correct outputs require interpreting law. Building on metamorphic testing, we introduce higher-order metamorphic relations that compare system outputs across structured shifts among similar individuals. Because authoring such relations is tedious and error-prone, we use an LLM-driven, role-based framework to automate test generation and code synthesis. We implement a multi-agent system that translates tax code into executable software and incorporates a metamorphic-testing agent that searches for counterexamples. In experiments, our framework using a smaller model (GPT-4o-mini) achieves a worst-case pass rate of 45%, outperforming frontier models (GPT-4o and Claude 3.5, 9-15%) on complex tax-code tasks. These results support agentic LLM methodologies as a path to robust, trustworthy legal-critical software from natural-language specifications.