Papers

Large Language Models (LLMs) are increasingly integrated into high-stakes applications, making robust safety guarantees a central practical and commercial concern. Existing safety evaluations predominantly rely on fixed collections of harmful prompts, implicitly assuming non-adaptive adversaries and thereby overlooking realistic attack scenarios in which inputs are iteratively refined to evade safeguards. In this work, we examine the vulnerability of contemporary language models to automated, adversarial prompt refinement. We repurpose black-box prompt optimization techniques, originally designed to improve performance on benign tasks, to systematically search for safety failures. Using DSPy, we apply three such optimizers to prompts drawn from HarmfulQA and JailbreakBench, explicitly optimizing toward a continuous danger score in the range 0 to 1 provided by an independent evaluator model (GPT-5.1). Our results demonstrate a substantial reduction in effective safety safeguards, with the effects being especially pronounced for open-source small language models. For example, the average danger score of Qwen 3 8B increases from 0.09 in its baseline setting to 0.79 after optimization. These findings suggest that static benchmarks may underestimate residual risk, indicating that automated, adaptive red-teaming is a necessary component of robust safety evaluation.

Unlearning knowledge is a pressing and challenging task in Large Language Models (LLMs) because of their unprecedented capability to memorize and digest training data at scale, raising more significant issues regarding safety, privacy, and intellectual property. However, existing works, including parameter editing, fine-tuning, and distillation-based methods, are all focused on flat sentence-level data but overlook the relational, multi-hop, and reasoned knowledge in naturally structured data. In response to this gap, this paper introduces Graph Oblivion and Node Erasure (GONE), a benchmark for evaluating knowledge unlearning over structured knowledge graph (KG) facts in LLMs.This KG-based benchmark enables the disentanglement of three effects of unlearning: direct fact removal, reasoning-based leakage, and catastrophic forgetting. In addition, Neighborhood-Expanded Distribution Shaping (NEDS), a novel unlearning framework, is designed to leverage graph connectivity and identify anchor-correlated neighbors, thereby enforcing a precise semantic separation between the forgotten fact and its semantic neighborhood. Evaluations on LLaMA-3-8B and Mistral-7B across multiple knowledge editing and unlearning methods showcase NEDS's superior performance (1.000 on unlearning efficacy and 0.839 on locality) on GONE and other benchmarks. The dataset is available at https://huggingface.co/datasets/GONE-Anonymous/GONE.

Dynamic runtime latency and memory constraints necessitate flexible large language model (LLM) deployment, where an LLM can be inferred with various quantization precisions based on available computational resources. Recent work on such any-precision quantization either relies on hardware-inefficient vector quantization or induces additional scaling factors when switching between bit-widths. Meanwhile, existing post-training quantization (PTQ) methods calibrated for a fixed low precision show poor generalizability under runtime precision change. In this work, we attribute the source of poor generalization across bit-widths to a precision-dependent outlier migration phenomenon where the distribution of PTQ-sensitive tokens changes across precisions. Motivated by this observation, we propose MoBiQuant, a novel any-precision Mixture-of-Bits quantization framework that adjusts weight precision for flexible LLM inference based on token sensitivity. Specifically, we propose a many-in-one recursive residual quantization that can iteratively reconstruct higher-precision weights at runtime and mitigates outlier migration with a token-aware router to dynamically select the optimal inference precision of each token.Extensive experiments show that MoBiQuant matches or surpasses frontier single-precision PTQ while exhibiting strong elasticity, achieving significant memory savings and throughput gains of up to $1.34\times$ over state-of-the-art any-precision methods. Our code is available at https://github.com/SimWangArizona/MoBiQuant.

Domain-specific speech remains a persistent challenge for automatic speech recognition (ASR), even for state-of-the-art systems like OpenAI's Whisper. We introduce Whisper: Courtside Edition, a novel multi-agent large language model (LLM) pipeline that enhances Whisper transcriptions without retraining. The pipeline intercepts Whisper's initial transcript, applies specialized LLM agents for domain context identification, named entity recognition, and jargon detection, and generates compact prompts that guide Whisper's decoder. Evaluated on 421 NBA basketball commentary segments (a domain characterized by dense proper nouns and technical terminology) our best pipeline achieves a statistically significant 17.0% relative reduction in word error rate (WER; from 0.217 to 0.180, p<0.001). Improvements are observed in 40.1% of segments with degradation in only 7.1%, substantially outperforming direct transcript post-editing. These results demonstrate that prompt-based augmentation can deliver scalable domain adaptation for ASR, offering a practical alternative to costly model fine-tuning.

Sarcasm detection poses a fundamental challenge in computational semantics, requiring models to resolve disparities between literal and intended meaning. The challenge is amplified in low-resource languages where annotated datasets are scarce or nonexistent. We present Yor-Sarc, the first gold-standard dataset for sarcasm detection in Yorùbá, a tonal Niger-Congo language spoken by over $50$ million people. The dataset comprises 436 instances annotated by three native speakers from diverse dialectal backgrounds using an annotation protocol specifically designed for Yorùbá sarcasm by taking culture into account. This protocol incorporates context-sensitive interpretation and community-informed guidelines and is accompanied by a comprehensive analysis of inter-annotator agreement to support replication in other African languages. Substantial to almost perfect agreement was achieved (Fleiss' $\kappa = 0.7660$; pairwise Cohen's $\kappa = 0.6732$–$0.8743$), with $83.3\%$ unanimous consensus. One annotator pair achieved almost perfect agreement ($\kappa = 0.8743$; $93.8\%$ raw agreement), exceeding a number of reported benchmarks for English sarcasm research works. The remaining $16.7\%$ majority-agreement cases are preserved as soft labels for uncertainty-aware modelling. Yor-Sarc\footnote{https://github.com/toheebadura/yor-sarc} is expected to facilitate research on semantic interpretation and culturally informed NLP for low-resource African languages.

Personal AI agents incur substantial cost via repeated LLM calls. We show existing caching methods fail: GPTCache achieves 37.9% accuracy on real benchmarks; APC achieves 0-12%. The root cause is optimizing for the wrong property – cache effectiveness requires key consistency and precision, not classification accuracy. We observe cache-key evaluation reduces to clustering evaluation and apply V-measure decomposition to separate these on n=8,682 points across MASSIVE, BANKING77, CLINC150, and NyayaBench v2, our new 8,514-entry multilingual agentic dataset (528 intents, 20 W5H2 classes, 63 languages). We introduce W5H2, a structured intent decomposition framework. Using SetFit with 8 examples per class, W5H2 achieves 91.1%+/-1.7% on MASSIVE in ~2ms – vs 37.9% for GPTCache and 68.8% for a 20B-parameter LLM at 3,447ms. On NyayaBench v2 (20 classes), SetFit achieves 55.3%, with cross-lingual transfer across 30 languages. Our five-tier cascade handles 85% of interactions locally, projecting 97.5% cost reduction. We provide risk-controlled selective prediction guarantees via RCPS with nine bound families.

The application of Large Language Models (LLMs) in accelerating scientific discovery has garnered increasing attention, with a key focus on constructing research agents endowed with innovative capability, i.e., the ability to autonomously generate novel and significant research ideas. Existing approaches predominantly rely on sophisticated prompt engineering and lack a systematic training paradigm. To address this, we propose DeepInnovator, a training framework designed to trigger the innovative capability of LLMs. Our approach comprises two core components. (1) “Standing on the shoulders of giants”. We construct an automated data extraction pipeline to extract and organize structured research knowledge from a vast corpus of unlabeled scientific literature. (2) “Conjectures and refutations”. We introduce a “Next Idea Prediction” training paradigm, which models the generation of research ideas as an iterative process of continuously predicting, evaluating, and refining plausible and novel next idea. Both automatic and expert evaluations demonstrate that our DeepInnovator-14B significantly outperforms untrained baselines, achieving win rates of 80.53%-93.81%, and attains performance comparable to that of current leading LLMs. This work provides a scalable training pathway toward building research agents with genuine, originative innovative capability, and will open-source the dataset to foster community advancement. Source code and data are available at: https://github.com/HKUDS/DeepInnovator.

Adeno-associated viruses (AAVs) are promising vectors for gene therapy, but their native serotypes face limitations in tissue tropism, immune evasion, and production efficiency. Engineering capsids to overcome these hurdles is challenging due to the vast sequence space and the difficulty of simultaneously optimizing multiple functional properties. The complexity also adds when it comes to the kidney, which presents unique anatomical barriers and cellular targets that require precise and efficient vector engineering. Here, we present AAVGen, a generative artificial intelligence framework for de novo design of AAV capsids with enhanced multi-trait profiles. AAVGen integrates a protein language model (PLM) with supervised fine-tuning (SFT) and a reinforcement learning technique termed Group Sequence Policy Optimization (GSPO). The model is guided by a composite reward signal derived from three ESM-2-based regression predictors, each trained to predict a key property: production fitness, kidney tropism, and thermostability. Our results demonstrate that AAVGen produces a diverse library of novel VP1 protein sequences. In silico validations revealed that the majority of the generated variants have superior performance across all three employed indices, indicating successful multi-objective optimization. Furthermore, structural analysis via AlphaFold3 confirms that the generated sequences preserve the canonical capsid folding despite sequence diversification. AAVGen establishes a foundation for data-driven viral vector engineering, accelerating the development of next-generation AAV vectors with tailored functional characteristics.

Large language models (LLMs) have demonstrated strong capabilities in complex reasoning tasks, yet their decision-making processes remain difficult to interpret. Existing explanation methods often lack trustworthy structural insight and are limited to single-instance analysis, failing to reveal reasoning stability and systematic failure mechanisms. To address these limitations, we propose the Trustworthy Unified Explanation Framework (TRUE), which integrates executable reasoning verification, feasible-region directed acyclic graph (DAG) modeling, and causal failure mode analysis. At the instance level, we redefine reasoning traces as executable process specifications and introduce blind execution verification to assess operational validity. At the local structural level, we construct feasible-region DAGs via structure-consistent perturbations, enabling explicit characterization of reasoning stability and the executable region in the local input space. At the class level, we introduce a causal failure mode analysis method that identifies recurring structural failure patterns and quantifies their causal influence using Shapley values. Extensive experiments across multiple reasoning benchmarks demonstrate that the proposed framework provides multi-level, verifiable explanations, including executable reasoning structures for individual instances, feasible-region representations for neighboring inputs, and interpretable failure modes with quantified importance at the class level. These results establish a unified and principled paradigm for improving the interpretability and reliability of LLM reasoning systems.

Self-supervised speech models (S3Ms) are known to encode rich phonetic information, yet how this information is structured remains underexplored. We conduct a comprehensive study across 96 languages to analyze the underlying structure of S3M representations, with particular attention to phonological vectors. We first show that there exist linear directions within the model's representation space that correspond to phonological features. We further demonstrate that the scale of these phonological vectors correlate to the degree of acoustic realization of their corresponding phonological features in a continuous manner. For example, the difference between [d] and [t] yields a voicing vector: adding this vector to [p] produces [b], while scaling it results in a continuum of voicing. Together, these findings indicate that S3Ms encode speech using phonologically interpretable and compositional vectors, demonstrating phonological vector arithmetic. All code and interactive demos are available at https://github.com/juice500ml/phonetic-arithmetic .

Robust and comprehensive evaluation of large language models (LLMs) is essential for identifying effective LLM system configurations and mitigating risks associated with deploying LLMs in sensitive domains. However, traditional statistical metrics are poorly suited to open-ended generation tasks, leading to growing reliance on LLM-based evaluation methods. These methods, while often more flexible, introduce additional complexity: they depend on carefully chosen models, prompts, parameters, and evaluation strategies, making the evaluation process prone to misconfiguration and bias. In this work, we present EvalSense, a flexible, extensible framework for constructing domain-specific evaluation suites for LLMs. EvalSense provides out-of-the-box support for a broad range of model providers and evaluation strategies, and assists users in selecting and deploying suitable evaluation methods for their specific use-cases. This is achieved through two unique components: (1) an interactive guide aiding users in evaluation method selection and (2) automated meta-evaluation tools that assess the reliability of different evaluation approaches using perturbed data. We demonstrate the effectiveness of EvalSense in a case study involving the generation of clinical notes from unstructured doctor-patient dialogues, using a popular open dataset. All code, documentation, and assets associated with EvalSense are open-source and publicly available at https://github.com/nhsengland/evalsense.

Large Language Models (LLMs) demonstrate strong reasoning performance, yet their ability to reliably monitor, diagnose, and correct their own errors remains limited. We introduce a psychologically grounded metacognitive framework that operationalizes Ann Brown's regulatory cycle (Planning, Monitoring, and Evaluation) as a structured prompting architecture, and study its integration within a lightweight dual-process MetaController for adaptive effort allocation. Across diverse reasoning and diagnostic benchmarks (GSM8K, CRUXEval, MBPP, AIME, CorrectBench, and TruthfulQA) using Llama-3 and Qwen-3 (8B), explicit regulatory structuring substantially improves error diagnosis and yields a threefold increase in successful self-correction. Blinded human evaluations over 580 query pairs show an 84% aggregate preference for trustworthiness and metacognitive self-awareness over standard and Chain-of-Thought baselines. Grounding LLM reasoning in established cognitive theory offers a principled path toward more transparent and diagnostically robust AI systems.

We introduce BURMESE-SAN, the first holistic benchmark that systematically evaluates large language models (LLMs) for Burmese across three core NLP competencies: understanding (NLU), reasoning (NLR), and generation (NLG). BURMESE-SAN consolidates seven subtasks spanning these competencies, including Question Answering, Sentiment Analysis, Toxicity Detection, Causal Reasoning, Natural Language Inference, Abstractive Summarization, and Machine Translation, several of which were previously unavailable for Burmese. The benchmark is constructed through a rigorous native-speaker-driven process to ensure linguistic naturalness, fluency, and cultural authenticity while minimizing translation-induced artifacts. We conduct a large-scale evaluation of both open-weight and commercial LLMs to examine challenges in Burmese modeling arising from limited pretraining coverage, rich morphology, and syntactic variation. Our results show that Burmese performance depends more on architectural design, language representation, and instruction tuning than on model scale alone. In particular, Southeast Asia regional fine-tuning and newer model generations yield substantial gains. Finally, we release BURMESE-SAN as a public leaderboard to support systematic evaluation and sustained progress in Burmese and other low-resource languages. https://leaderboard.sea-lion.ai/detailed/MY

Defending LLMs against adversarial jailbreak attacks remains an open challenge. Existing defenses rely on binary classifiers that fail when adversarial input falls outside the learned decision boundary, and repeated fine-tuning is computationally expensive while potentially degrading model capabilities. We propose MANATEE, an inference-time defense that uses density estimation over a benign representation manifold. MANATEE learns the score function of benign hidden states and uses diffusion to project anomalous representations toward safe regions–requiring no harmful training data and no architectural modifications. Experiments across Mistral-7B-Instruct, Llama-3.1-8B-Instruct, and Gemma-2-9B-it demonstrate that MANATEE reduce Attack Success Rate by up to 100% on certain datasets, while preserving model utility on benign inputs.

We present ArabicNumBench, a comprehensive benchmark for evaluating large language models on Arabic number reading tasks across Eastern Arabic-Indic numerals (0-9 in Arabic script) and Western Arabic numerals (0-9). We evaluate 71 models from 10 providers using four prompting strategies (zero-shot, zero-shot CoT, few-shot, few-shot CoT) on 210 number reading tasks spanning six contextual categories: pure numerals, addresses, dates, quantities, and prices. Our evaluation comprises 59,010 individual test cases and tracks extraction methods to measure structured output generation. Evaluation reveals substantial performance variation, with accuracy ranging from 14.29% to 99.05% across models and strategies. Few-shot Chain-of-Thought prompting achieves 2.8x higher accuracy than zero-shot approaches (80.06% vs 28.76%). A striking finding emerges: models achieving elite accuracy (98-99%) often produce predominantly unstructured output, with most responses lacking Arabic CoT markers. Only 6 models consistently generate structured output across all test cases, while the majority require fallback extraction methods despite high numerical accuracy. Comprehensive evaluation of 281 model-strategy combinations demonstrates that numerical accuracy and instruction-following represent distinct capabilities, establishing baselines for Arabic number comprehension and providing actionable guidance for model selection in production Arabic NLP systems.

This paper establishes a fundamental convergence: Schema-Guided Dialogue (SGD) and the Model Context Protocol (MCP) represent two manifestations of a unified paradigm for deterministic, auditable LLM-agent interaction. SGD, designed for dialogue-based API discovery (2019), and MCP, now the de facto standard for LLM-tool integration, share the same core insight – that schemas can encode not just tool signatures but operational constraints and reasoning guidance. By analyzing this convergence, we extract five foundational principles for schema design: (1) Semantic Completeness over Syntactic Precision, (2) Explicit Action Boundaries, (3) Failure Mode Documentation, (4) Progressive Disclosure Compatibility, and (5) Inter-Tool Relationship Declaration. These principles reveal three novel insights: first, SGD's original design was fundamentally sound and should be inherited by MCP; second, both frameworks leave failure modes and inter-tool relationships unexploited – gaps we identify and resolve; third, progressive disclosure emerges as a critical production-scaling insight under real-world token constraints. We provide concrete design patterns for each principle. These principles position schema-driven governance as a scalable mechanism for AI system oversight without requiring proprietary system inspection – central to Software 3.0.

Retrieval-Augmented Generation (RAG) has demonstrated strong effectiveness in knowledge-intensive tasks by grounding language generation in external evidence. Despite its success, many existing RAG systems are built based on a ranking-centric, asymmetric dependency paradigm, where the generation quality of the generator is highly dependent on reranking results of the reranker. To overcome this limitation, we propose Cooperative Retrieval-Augmented Generation (CoRAG), a framework that treats the reranker and the generator as peer decision-makers rather than being connected through an asymmetric dependency pipeline. By jointly optimizing their behaviors toward a shared task objective, the reranker and generator are encouraged to cooperate, ensuring that document reranking and generation work in concert to improve the final response. Experimental results demonstrate good generalization and improved generation stability of CoRAG, even when the model is trained on only around 10K PopQA samples. Our model released in https://github.com/CoderrrSong/CoRAG.

Semi-supervised learning in automatic speech recognition (ASR) typically relies on pseudo-labeling, which often suffers from confirmation bias and error accumulation due to noisy supervision. To address this limitation, we propose ReHear, a framework for iterative pseudo-label refinement that integrates an instruction-tuned, audio-aware large language model (LLM) into the self-training loop. Unlike conventional text-based correctors, our approach conditions the LLM on both the ASR hypothesis and the source audio, allowing it to recover phonetically accurate transcripts even from severe recognition errors. These refined pseudo-labels serve as high-fidelity targets for fine-tuning the ASR model in an iterative cycle. Experimental results across diverse benchmarks demonstrate that ReHear effectively mitigates error propagation and improves over both supervised and pseudo-labeling baselines.

LLM agents rely heavily on high-quality trajectory data to guide their problem-solving behaviors, yet producing such data requires substantial task design, high-capacity model generation, and manual filtering. Despite the high cost of creating these datasets, existing literature has overlooked copyright protection for LLM agent trajectories. This gap leaves creators vulnerable to data theft and makes it difficult to trace misuse or enforce ownership rights. This paper introduces ActHook, the first watermarking method tailored for agent trajectory datasets. Inspired by hook mechanisms in software engineering, ActHook embeds hook actions that are activated by a secret input key and do not alter the original task outcome. Like software execution, LLM agents operate sequentially, allowing hook actions to be inserted at decision points without disrupting task flow. When the activation key is present, an LLM agent trained on watermarked trajectories can produce these hook actions at a significantly higher rate, enabling reliable black-box detection. Experiments on mathematical reasoning, web searching, and software engineering agents show that ActHook achieves an average detection AUC of 94.3 on Qwen-2.5-Coder-7B while incurring negligible performance degradation.

Studies of semantic drift report heterogeneous signals, including embedding displacement, neighbor change, distributional divergence, and recursive trajectory instability, without a shared account that relates them. Semantic Substrate Dynamics Theory (SSDT) treats these signals as observables of one time-indexed substrate, St = (X, dt, Pt), that couples embedding geometry to a local diffusion kernel. The contribution is commensurability with a mechanism layer: the substrate separates within-basin churn from basin crossing, recursion-induced instability, and intervention-order effects, distinctions that a single detection score does not recover. Coarse Ricci curvature functions as a dense structural descriptor of basin and bridge geometry across the graph, and bridge mass, a node-level aggregate of incident negative curvature, functions as a sparse descriptor of the genuine bridge structure that is typically uncommon in embedding graphs. For recursive generation, node displacement relative to an origin decomposes into a radial component and a tangential component, which separates bounded departure from continuing reinterpretation. The predictions are stated in falsifiable form with a pre-declared rejection rule, and the predicted leading indicator of future rewiring is a local density statistic rather than the curvature aggregate. This manuscript provides the formal model, the assumptions, the observable roles, and the test contracts; empirical performance is deferred.

The spread of misinformation across digital platforms can pose significant societal risks. Claim verification, a.k.a. fact-checking, systems can help identify potential misinformation. However, their efficacy is limited by the knowledge sources that they rely on. Most automated claim verification systems depend on a single knowledge source and utilize the supporting evidence from that source; they ignore the disagreement of their source with others. This limits their knowledge coverage and transparency. To address these limitations, we present a novel system for open-domain claim verification (ODCV) that leverages large language models (LLMs), multi-perspective evidence retrieval, and cross-source disagreement analysis. Our approach introduces a novel retrieval strategy that collects evidence for both the original and the negated forms of a claim, enabling the system to capture supporting and contradicting information from diverse sources: Wikipedia, PubMed, and Google. These evidence sets are filtered, deduplicated, and aggregated across sources to form a unified and enriched knowledge base that better reflects the complexity of real-world information. This aggregated evidence is then used for claim verification using LLMs. We further enhance interpretability by analyzing model confidence scores to quantify and visualize inter-source disagreement. Through extensive evaluation on four benchmark datasets with five LLMs, we show that knowledge aggregation not only improves claim verification but also reveals differences in source-specific reasoning. Our findings underscore the importance of embracing diversity, contradiction, and aggregation in evidence for building reliable and transparent claim verification systems

Anxiety is the unease about a possible future negative outcome. In recent years, there has been growing interest in understanding how anxiety relates to our health, well-being, body, mind, and behaviour. This includes work on lexical resources for word-anxiety association. However, there is very little anxiety-related work on larger units of text such as multiword expressions (MWE). Here, we introduce the first large-scale lexicon capturing descriptive norms of anxiety associations for more than 20k English MWEs. We show that the anxiety associations are highly reliable. We use the lexicon to study prevalence of different types of anxiety- and calmness-associated MWEs; and how that varies across two-, three-, and four-word sequences. We also study the extent to which the anxiety association of MWEs is compositional (due to its constituent words). The lexicon enables a wide variety of anxiety-related research in psychology, NLP, public health, and social sciences. The lexicon is freely available: https://saifmohammad.com/worrylex.html

We reinterpret the final Large Language Model (LLM) softmax classifier as an Energy-Based Model (EBM), decomposing the sequence-to-sequence probability chain into multiple interacting EBMs at inference. This principled approach allows us to track "energy spills" during decoding, which we empirically show correlate with factual errors, biases, and failures. Similar to Orgad et al. (2025), our method localizes the exact answer token and subsequently tests for hallucinations. Crucially, however, we achieve this without requiring trained probe classifiers or activation ablations. Instead, we introduce two completely training-free metrics derived directly from output logits: spilled energy, which captures the discrepancy between energy values across consecutive generation steps that should theoretically match, and marginalized energy, which is measurable at a single step. Evaluated on nine benchmarks across state-of-the-art LLMs (including LLaMA, Mistral, and Gemma) and on synthetic algebraic operations (Qwen3), our approach demonstrates robust, competitive hallucination detection and cross-task generalization. Notably, these results hold for both pretrained and instruction-tuned variants without introducing any training overhead. Code available at: github.com/OmnAI-Lab/spilled-energy

We introduce ACE-TA, the Agentic Coding and Explanations Teaching Assistant framework, that autonomously routes conceptual queries drawn from programming course material to grounded Q&A, stepwise coding guidance, and automated quiz generation using pre-trained Large Language Models (LLMs). ACE-TA consists of three coordinated modules: a retrieval grounded conceptual Q&A system that provides precise, context-aligned explanations; a quiz generator that constructs adaptive, multi-topic assessments targeting higher-order understanding; and an interactive code tutor that guides students through step-by-step reasoning with sandboxed execution and iterative feedback.

High-fidelity diagram creation requires coordinated visual-spatial decisions over semantic topology, visual styling, and geometric layout. Existing methods face a representation gap: pixel-based generation offers limited object-level control, while code-based synthesis provides executable structure at the cost of intuitive manipulation. We introduce EvoDesign, an agentic framework for editable diagram creation through an intermediate canvas schema that is both machine-actionable and directly manipulable by users. EvoDesign first constructs a diagram manifest through coordinated structure, style, and layout agents, and then renders the manifest in a diagnostic verification environment. This environment produces objective defect signals and VLM-based critique, enabling localized refinement of the current diagram while also providing evidence for long-term design expertise evolution. The evolution mechanism distills refinement traces into a three-tier design expertise memory, where candidate strategies are linked to the verification evidence that supports them and are promoted into broader guidelines and principles only through repeated cross-context support. We further define CanvasBench, a canvas-recoverable benchmark and evaluation protocol with content, visual, and cognitive dimensions. Experiments across pixel, code-based, agentic, one-shot schema, and canvas-native methods show that EvoDesign provides the strongest overall balance across content fidelity, visual organization, and cognitive usability. Our code is available at https://github.com/GeminiLight/EvoDesign.

People increasingly hold sustained, open-ended conversations with large language models (LLMs). Public reports and early studies suggest that, in such settings, models can reinforce delusional or conspiratorial ideation or even amplify harmful beliefs and engagement patterns. We present an audit and benchmarking study that measures how different LLMs encourage, resist, or escalate disordered and conspiratorial thinking. We explicitly compare API outputs to user chat interfaces, like the ChatGPT desktop app or web interface, which is how people have conversations with chatbots in real life but are almost never used for testing. In total, we run 56 20-turn conversations testing ChatGPT-4o and ChatGPT-5, via both the API and chat interface, and grade each conversation by two research assistants (RAs) as well as by GPT-5. We document five results. First, we observe large differences in performance between the API and chat interface environments, showing that the universally used method of automated testing through the API is not sufficient to assess the impact of chatbots in the real world. Second, when tested in the chat interface, we find that ChatGPT-5 displays less sycophancy, escalation, and delusion reinforcement than ChatGPT-4o, showing that these behaviors are influenced by the policy choices of major AI companies. Third, conversations with nearly identical aggregate intensity in a behavior display large differences in how the behavior evolves turn by turn, highlighting the importance of temporal dynamics in multi-turn evaluation. Fourth, even updated models display substantial levels of negative behaviors, revealing that model improvement does not imply model safety. Fifth, the same API endpoint tested just two months apart yields a complete reversal in behavior, underscoring how transparency in model updates is a necessary prerequisite for robust audit findings.

The study explores whether current Large Language Models (LLMs) exhibit Theory of Mind (ToM) capabilities – specifically, the ability to infer others' beliefs, intentions, and emotions from text. Given that LLMs are trained on language data without social embodiment or access to other manifestations of mental representations, their apparent social-cognitive reasoning raises key questions about the nature of their understanding. Are they capable of robust mental-state attribution indistinguishable from human ability in its output, or do their outputs merely reflect superficial pattern completion? To address this question, we tested five LLMs and compared their performance to that of human controls using an adapted version of a text-based tool widely used in human ToM research. The test involves answering questions about the beliefs, intentions, and emotions of story characters. The results revealed a performance gap between the models. Earlier and smaller models were strongly affected by the number of relevant inferential cues available and, to some extent, were also vulnerable to the presence of irrelevant or distracting information in the texts. In contrast, GPT-4o demonstrated high accuracy and strong robustness, performing comparably to humans even in the most challenging conditions. This work contributes to ongoing debates about the cognitive status of LLMs and the boundary between genuine understanding and statistical approximation.

This thesis develops a system for automatically analyzing and improving dynamic programs, such as those that have driven progress in natural language processing and computer science, more generally, for decades. Finding a correct program with the optimal asymptotic runtime can be unintuitive, time-consuming, and error-prone. This thesis aims to automate this laborious process. To this end, we develop an approach based on 1. a high-level, domain-specific language called Dyna for concisely specifying dynamic programs 2. a general-purpose solver to efficiently execute these programs 3. a static analysis system that provides type analysis and worst-case time/space complexity analyses 4. a rich collection of meaning-preserving transformations to programs, which systematizes the repeated insights of numerous authors when speeding up algorithms in the literature 5. a search algorithm for identifying a good sequence of transformations that reduce the runtime complexity, given an initial, correct program We show that, in practice, automated search – like the mental search performed by human programmers – can find substantial improvements to the initial program. Empirically, we show that many speed-ups described in the NLP literature could have been discovered automatically by our system. We provide a freely available prototype system at https://github.com/timvieira/dyna-pi.

We present an end-to-end self-hosted (API-free) pipeline, where API-free means that lecture content is not sent to any external LLM service, that converts lecture PDFs into multiple-choice questions (MCQs) using a local LLM plus deterministic quality control (QC). The pipeline is designed for black-box minimization: LLMs may assist drafting, but the final released artifacts are plain-text question banks with an explicit QC trace and without any need to call an LLM at deployment time. We run a seed sweep on three short "dummy lectures" (information theory, thermodynamics, and statistical mechanics), collecting 15 runs x 8 questions = 120 accepted candidates (122 attempts total under bounded retries). All 120 accepted candidates satisfy hard QC checks (JSON schema conformance, a single marked correct option, and numeric/constant equivalence tests); however, the warning layer flags 8/120 items (spanning 8 runs) that expose residual quality risks such as duplicated distractors or missing rounding instructions. We report a warning taxonomy with concrete before->after fixes, and we release the final 24-question set (three lectures x 8 questions) as JSONL/CSV for Google Forms import (e.g., via Apps Script or API tooling) included as ancillary files under anc/. Finally, we position the work through the AI to Learn (AI2L) rubric lens and argue that self-hosted MCQ generation with explicit QC supports privacy, accountability, and Green AI in educational workflows.

The growing demand for automated graph algorithm reasoning has attracted increasing attention in the large language model (LLM) community. Recent LLM-based graph reasoning methods typically decouple task descriptions from graph data, generate executable code augmented by retrieval from technical documentation, and refine the code through debugging. However, we identify two key limitations in existing approaches: (i) they treat technical documentation as flat text collections and ignore its hierarchical structure, leading to noisy retrieval that degrades code generation quality; and (ii) their debugging mechanisms focus primarily on runtime errors, yet ignore more critical logical errors. To address them, we propose {\method}, an agentic hierarchical retrieval-augmented coding framework that exploits the document hierarchy through top-down traversal and early pruning, together with a self-debugging coding agent that iteratively refines code using automatically generated small-scale test cases. To enable comprehensive evaluation of complex graph reasoning, we introduce a new dataset, {\dataset}, covering small-scale, large-scale, and composite graph reasoning tasks. Extensive experiments demonstrate that our method achieves higher task accuracy and lower inference cost compared to baselines\footnote{The code is available at \href{https://github.com/FairyFali/GraphSkill}{\textcolor{blue}{https://github.com/FairyFali/GraphSkill}}.}.