arXiv subjects

Giuseppe Carenini

Publications and source records attributed to Giuseppe Carenini.

At least 19 recordsLinked to original sources

arXiv2026

UnpredictaBench: A Benchmark for Evaluating Distributional Randomness in LLMs

We introduce UnpredictaBench, an evaluation that tests the ability of large language models (LLMs) to capture true underlying distributions. As LLMs are increasingly used as substitutes for other entities (e.g., for humans in economic simulations), the tendency of many models to collapse towards a single plausible answer means a failure to capture the unpredictability of real systems. Recent work on improving output diversity is insufficient for this setting: simulation requires samples that are calibrated to a target distribution, not merely varied outputs. UnpredictaBench isolates a simplified but fundamental version of this problem: sampling outcomes from individual target distributions, including canonical statistical distributions, distributions induced by stochastic programs, and natural-language scenarios that describe random processes. We introduce 448 such problems together with KS@N, a general-purpose evaluation metric that quantifies how well a model outputs approximate black-box target distributions via the Kolmogorov-Smirnov statistical test. This is the rate at which we fail to reject model samples of size N against ground-truth samples, with larger N indicating greater difficulty. Tested across open and proprietary models, we find a large spread in distributional capabilities. For instance, when models generate samples of size 100 (KS@100, our standard metric), scores range from near 0 to over 20%. No model is able to achieve over 40% at KS@100, showing significant headroom in distributional sampling as a capability. Although adding reasoning can somewhat increase scores, we find no immediate solution for this issue. UnpredictaBench shows that even simple distributional simulation remains challenging, making it a necessary first step toward using LLMs as stand-ins for complex systems. Project website and resources are available at https://unpredictabenchmark.github.io/.

cs.CL↗

arXiv2026

Dual-Confidence Contrastive Decoding for Retrieval-Augmented Generation

Retrieval-augmented generation (RAG) increasingly requires models to answer questions from multiple retrieved documents, where only some sources are relevant and the retrieved bundle may contain stale, noisy, or conflicting evidence. Existing contrastive decoding methods primarily focus on resolving conflicts between the model's internal memory and the retrieved context. In contrast, we study the complementary problem of intra-context conflict in multi-document RAG. To evaluate this setting, we introduce DRQA, a factual-conflict question answering benchmark derived from enterprise deep-research scenarios, where answers are grounded in synthetic enterprise-specific facts that are designed not to be recoverable from the model's internal memory. We further propose Dual-Confidence Contrastive Decoding (DCCD), a training-free decoding method that combines document-level confidence, which estimates whether a document appears sufficient for answering the question, with token-level confidence, which estimates whether that document supports a confident next-token prediction. DCCD selects positive and negative document-conditioned streams using these dual-confidence signals and scales a document-level contrast by their confidence margin. Across DRQA and standard multi-document QA benchmarks, DCCD achieves the best average performance among full-context and contrastive decoding baselines, with the largest gains on DRQA. These results highlight the importance of source-aware, confidence-gated decoding when retrieved evidence is internally conflicting.

cs.CL↗

arXiv2026

SeKV: Resolution-Adaptive KV Cache with Hierarchical Semantic Memory for Long-Context LLM Inference

Large language models increasingly operate over long contexts, where the KV cache becomes a dominant memory bottleneck: its size grows linearly with sequence length and must be retained throughout decoding, making full GPU caching prohibitively expensive without compression. Existing KV cache compression methods struggle to balance efficiency with faithful context preservation. Token eviction discards information, while semantic grouping fixes compression decisions at prefill time; neither can recover token-level detail from a compressed span once it becomes relevant during generation. As a solution, we propose SeKV, a resolution-adaptive semantic KV cache that organizes context into entropy-guided semantic spans and stores them across a GPU-CPU memory hierarchy without discarding information. Each span keeps a lightweight summary vector on GPU for coarse routing and a low-rank SVD basis on CPU for on-demand token-level reconstruction. A trained zoom-in mechanism selectively expands query-relevant spans during decoding, enabling precise retrieval without materializing the full KV cache on GPU. SeKV enables adaptive token-level reconstruction while keeping the base LLM fully frozen and adding fewer than 0.05% trainable parameters. Across four benchmarks, SeKV improves over the strongest semantic compression baseline by 5.9% on average while reducing GPU memory by 53.3% versus full KV caching at 128K context. Code is available on https://github.com/AmirAbaskohi/SeKV.

cs.CL↗

arXiv2026

SproutRAG: Attention-Guided Tree Search with Progressive Embeddings for Long-Document RAG

Retrieval-augmented generation (RAG) systems must balance retrieval granularity with contextual coherence, a challenge that existing methods address through LLM-guided chunking, single-level context expansion, or hierarchical summarization. These approaches variously depend on costly LLM calls during indexing or retrieval, limit context aggregation to a single granularity level, or introduce information loss through summarization. We present SproutRAG, an attention-guided hierarchical RAG framework that addresses this trade-off by organizing sentence-level chunks into progressively larger but semantically coherent units, using learned inter-sentence attention to construct a binary chunking tree. Unlike prior approaches that rely on external LLMs, fixed context expansion, or lossy summarization, SproutRAG learns which attention heads and layers best capture semantic document structure, enabling multi-granularity retrieval without additional LLM calls or compressed summaries. At retrieval time, SproutRAG uses hierarchical beam search to retrieve candidates at multiple granularities, capturing multi-sentence relevance beyond flat retrieval. The framework is trained end-to-end with a joint objective that improves both embeddings and tree structure. Experiments across four benchmarks spanning scientific, legal, and open-domain settings demonstrate that SproutRAG improves information efficiency (IE) by 6.1% on average over the strongest baseline. Code is available on https://github.com/AmirAbaskohi/SproutRAG.

cs.CL↗

arXiv2026

MCompassRAG: Topic Metadata as a Semantic Compass for Paragraph-Level Retrieval

Retrieval-augmented generation (RAG) systems depend critically on how documents are chunked and searched. Fine-grained chunks can improve retrieval precision but expand the search space, increasing latency and cost; larger chunks reduce the number of candidates but make dense similarity less reliable, as the representation for each chunk mixes multiple topics and introduces more semantic noise. This trade-off becomes especially limiting in deep research tasks, where retrieval must be both fast and precise across large, heterogeneous corpora. We introduce MCompassRAG, a metadata-guided retrieval framework that uses topic-level signals as a semantic compass for selecting relevant evidence. Instead of relying only on cosine similarity between queries and noisy chunk embeddings, MCompassRAG enriches chunk representations with topic metadata in the same embedding space and trains a lightweight retriever through LLM-teacher distillation. At inference time, MCompassRAG performs topic-aware retrieval without additional LLM calls, improving both efficiency and evidence quality. Across six complex retrieval benchmarks, MCompassRAG improves information efficiency (IE) by 8.24% on average with over 5 times lower latency than the strongest efficient RAG baselines. Code is available on https://github.com/AmirAbaskohi/MCompassRAG.

cs.CL↗

arXiv2026

The Illusion of Multi-Agent Advantage

Prevailing wisdom posits that Multi-Agent Systems (MAS) are superior to Single-Agent Systems (SAS), citing advantages like context protection, parallel processing and distributed decision-making. However, empirical support for this claim relies primarily on comparisons with SAS baselines using benchmarks that prioritize isolated reasoning tasks, which do not adequately assess these advantages. Focusing on automatically generated MAS that are designed for enhanced generalizability over manually-designed counterparts, we perform a rigorous, systematic evaluation against SAS, specifically Chain-of-Thought with Self-Consistency (CoT-SC). Across traditional reasoning datasets and tasks with interactive multi-step workflows (e.g., BrowseComp-Plus), we demonstrate that automatic MAS consistently underperform CoT-SC despite being up to 10x more expensive. To isolate these failures from limitations inherent to task structure, we introduce a diagnostic synthetic dataset tailored for MAS featuring explicit task decomposition, context separation and parallelization potential. We show that expert-architected MAS consistently outperforms automatically generated architectures in both raw performance and cost-efficiency on this dataset, demonstrating that existing evaluation frameworks mask critical architectural gaps and inefficiencies of complex MAS by failing to account for the marginal utility of increased computational cost. Critically, systematic deconstruction of the generated MAS architectures reveals that current automated design paradigms produce architectural bloat that prioritizes superficial complexity which does not translate into functional utility, exposing a fundamental misalignment with multi-agent principles.

cs.AI↗

arXiv2026

Improving Topic Modeling by Distilling Soft Labels from Language Models

Traditional neural topic models are typically optimized by reconstructing the document's Bag-of-Words (BoW) representations, overlooking contextual information and struggling with data sparsity. In this work, we introduce a novel topic model training framework by Distilling Soft Labels (DSL) from Language Models (LMs). To construct the contextually enriched reconstruction signals, we project the next token probabilities, conditioned on a specialized prompt, onto a pre-defined vocabulary, and train the topic models to reconstruct the soft labels using the LM hidden states. This produces higher-quality topics that are more closely aligned with the underlying thematic structure of the corpus. Extensive experiments demonstrate that DSL achieves substantial improvements in topic coherence and assignment accuracy over existing baselines. Additionally, we also introduce a retrieval-based metric, which shows that our approach significantly outperforms existing methods in identifying semantically similar documents, highlighting its effectiveness for retrieval-oriented applications.

cs.CL↗

arXiv2026

ReVision: Scaling Computer-Use Agents via Temporal Visual Redundancy Reduction

Computer-use agents (CUAs) rely on visual observations of graphical user interfaces, where each screenshot is encoded into a large number of visual tokens. As interaction trajectories grow, the token cost increases rapidly, limiting the amount of history that can be incorporated under fixed context and compute budgets. This has resulted in no or very limited improvement in the performance when using history unlike other domains. We address this inefficiency by introducing ReVision, which is used to train multimodal language models on trajectories where redundant visual patches are removed using a learned patch selector that compares patch representations across consecutive screenshots while preserving spatial structure required by the model. Across three benchmarks, OSWorld, WebTailBench, and AgentNetBench, when processing trajectories with 5 history screenshots using Qwen2.5-VL-7B, ReVision reduces token usage by 46% on average while improving success rate by 3% over the no drop baseline. This establishes a clear efficiency gain, enabling agents to process longer trajectories with fewer tokens. With this improved efficiency, we revisit the role of history in CUAs and find that performance continues to improve as more past observations are incorporated when redundancy is removed.

cs.CL↗

arXiv2026

VAMPS: Visual-Assisted Mathematical Problem Solving Benchmark

Multimodal large language models are increasingly capable of complex reasoning, yet their performance often degrades when they must externalize a problem through a tool and then reason over the tool's output, specifically when they rely on visual aids. This gap is especially important because real engineering and scientific workflows often rely on visualization tools for analysis, validation, and decision-making. To study this discrepancy, we introduce VAMPS (Visual-Assisted Mathematical Problem Solving), a benchmark for graph-assisted mathematics. VAMPS contains 1,168 multimodal, bilingual multiple-choice question-answer pairs drawn from Iranian University Entrance Exam algebra and calculus problems and expanded with human-reviewed LLM-generated synthetic variants, all selected so that plotting provides a natural solution strategy by revealing intersections, extrema, asymptotes, etc. Designed for both benchmarking and diagnosis, VAMPS goes beyond prior multimodal benchmarks that primarily evaluate reasoning over fixed visual inputs by testing whether a model can benefit from constructing a useful graph and grounding its answer in the resulting visualization. Overall, we found that across a diverse set of models, direct analytical solving surprisingly outperforms tool-enabled visual solving, even on problems where plotting is a natural strategy.

cs.AI↗

arXiv2026

IntentGrasp: A Comprehensive Benchmark for Intent Understanding

Accurately understanding the intent behind speech, conversation, and writing is crucial to the development of helpful Large Language Model (LLM) assistants. This paper introduces IntentGrasp, a comprehensive benchmark for evaluating the intent understanding capability of LLMs. Derived from 49 high-quality, open-licensed corpora spanning 12 diverse domains, IntentGrasp is constructed through source datasets curation, intent label contextualization, and task format unification. IntentGrasp contains a large-scale training set of 262,759 instances and two evaluation sets: an All Set of 12,909 test cases and a more balanced and challenging Gem Set of 470 cases. Extensive evaluations on 20 LLMs across 7 families (including frontier models such as GPT-5.4, Gemini-3.1-Pro, and Claude-Opus-4.7) demonstrate unsatisfactory performance, with scores below 60% on All Set and below 25% on Gem set. Notably, 17 out of 20 tested models perform worse than a random-guess baseline (15.2%) on Gem Set, while the estimated human performance is ~81.1%, showing substantial room for improvement. To enhance such ability, this paper proposes Intentional Fine-Tuning (IFT), which fine-tunes the models on the training set in IntentGrasp, yielding significant gains of 30+ F1 points on All Set and 20+ points on Gem Set. Tellingly, the leave-one-domain-out (Lodo) experiments further demonstrate the strong cross-domain generalizability of IFT, verifying that it is a promising approach to substantially enhancing the intent understanding of LLMs. Overall, by benchmarking and boosting intent understanding ability, this study sheds light on a promising path towards more intentional, capable, and safe AI assistants for human benefits and social good.

cs.CL↗

arXiv2026

Improving Language Models with Intentional Analysis

Intent, a critical cognitive notion and mental state, is ubiquitous in human communication and problem-solving. Accurately understanding the underlying intent behind questions is imperative to reasoning towards correct answers. However, this significant concept has been largely disregarded in the rapid development of language models (LMs). To unleash the potential of intent and instill it into LMs, this paper introduces Intentional Analysis (IA), which explicitly invokes intent-aware analysis and reasoning during the problem-solving process. Comprehensive experiments across diverse benchmarks, model types, and configurations demonstrate the effectiveness, robustness, and generalizability of IA. Notably, IA consistently improves task performance even on SOTA proprietary models like GPT-5 and Claude-Opus-4.6. Moreover, IA not only outperforms Chain-of-Thought (CoT) across various experimental settings, but it can also synergistically work with CoT reasoning. Further qualitative analysis and case studies reveal that the benefits of IA stem from addressing several weaknesses in baseline methods, such as intent misunderstanding, hasty generalization, and mental laziness. Case studies also provide insights into the mechanisms underlying IA and clarify how it differs from CoT in mitigating these weaknesses. This study sheds light on a promising direction for the development of future LLMs with intentional analysis.

cs.CL↗

arXiv2026

Infusing Theory of Mind into Socially Intelligent LLM Agents

Theory of Mind (ToM)-an understanding of the mental states of others-is a key aspect of human social intelligence, yet, chatbots and LLM-based social agents do not typically integrate it. In this work, we demonstrate that LLMs that explicitly use ToM get better at dialogue, achieving goals more effectively. After showing that simply prompting models to generate mental states between dialogue turns already provides significant benefit, we further introduce ToMAgent (ToMA), a ToM-focused dialogue agent. ToMA is trained by pairing ToM with dialogue lookahead to produce mental states that are maximally useful for achieving dialogue goals. Experiments on the Sotopia interactive social evaluation benchmark demonstrate the effectiveness of our method over a range of baselines. Comprehensive analysis shows that ToMA exhibits more strategic, goal-oriented reasoning behaviors, which enable long-horizon adaptation, while maintaining better relationships with their partners. Our results suggest a step forward in integrating ToM for building socially intelligent LLM agents.

cs.CL↗

arXiv2026

When Minor Edits Matter: LLM-Driven Prompt Attack for Medical VLM Robustness in Ultrasound

Ultrasound is widely used in clinical practice due to its portability, cost-effectiveness, safety, and real-time imaging capabilities. However, image acquisition and interpretation remain highly operator dependent, motivating the development of robust AI-assisted analysis methods. Vision-language models (VLMs) have recently demonstrated strong multimodal reasoning capabilities and competitive performance in medical image analysis, including ultrasound. However, emerging evidence highlights significant concerns about their trustworthiness. In particular, adversarial robustness is critical because Med-VLMs operate via natural-language instructions, rendering prompt formulation a realistic and practically exploitable point of vulnerability. Small variations (typos, shorthand, underspecified requests, or ambiguous wording) can meaningfully shift model outputs. We propose a scalable adversarial evaluation framework that leverages a large language model (LLM) to generate clinically plausible adversarial prompt variants via "humanized" rewrites and minimal edits that mimic routine clinical communication. Using ultrasound multiple-choice question answering benchmarks, we systematically assess the vulnerability of SOTA Med-VLMs to these attacks, examine how attacker LLM capacity influences attack success, analyze the relationship between attack success and model confidence, and identify consistent failure patterns across models. Our results highlight realistic robustness gaps that must be addressed for safe clinical translation. Code will be released publicly following the review process.

cs.CV↗

arXiv2026

DRBench: A Realistic Benchmark for Enterprise Deep Research

We introduce DRBench, a benchmark for evaluating AI agents on complex, open-ended deep research tasks in enterprise settings. Unlike prior benchmarks that focus on simple questions or web-only queries, DRBench evaluates agents on multi-step queries (for example, "What changes should we make to our product roadmap to ensure compliance with this standard?") that require identifying supporting facts from both the public web and private company knowledge base. Each task is grounded in realistic user personas and enterprise context, spanning a heterogeneous search space that includes productivity software, cloud file systems, emails, chat conversations, and the open web. Tasks are generated through a carefully designed synthesis pipeline with human-in-the-loop verification, and agents are evaluated on their ability to recall relevant insights, maintain factual accuracy, and produce coherent, well-structured reports. We release 100 deep research tasks across 10 domains, such as Sales, Cybersecurity, and Compliance. We demonstrate the effectiveness of DRBench by evaluating diverse DR agents across open- and closed-source models (such as GPT, Llama, and Qwen) and DR strategies, highlighting their strengths, weaknesses, and the critical path for advancing enterprise deep research. Code and data are available at https://github.com/ServiceNow/drbench.

cs.CL↗

arXiv2026

SOP-Bench: Complex Industrial SOPs for Evaluating LLM Agents

LLM-based agents struggle to execute complex, multi-step Standard Operating Procedures (SOPs) that are fundamental to industrial automation. Existing benchmarks fail to capture the procedural complexity and tool orchestration demands of real-world workflows. We introduce SOP-Bench, a benchmark of 2,000+ tasks from human expert-authored SOPs across 12 business domains (healthcare, logistics, finance, content moderation, etc.). Using a human-AI collaborative framework, experts crafted authentic SOPs while AI generated artifacts (tools, APIs, datasets), all human-validated, yielding realistic tasks with executable interfaces and ground-truth outputs. SOP-Bench serves as a research enabler for systematically investigating agent architectures, model capabilities, and deployment considerations across diverse procedural tasks. We demonstrate its utility through illustrative experiments with a subset of frontier models across Function-Calling (FC) and ReAct agents, revealing critical insights. For example, (1) newer models do not guarantee better performance - Claude 4 family outperforms Claude 4.5 family on ReAct tasks (Claude 4 Opus: 72.4% vs. Claude 4.5 Sonnet: 63.3% task success rate), demonstrating that production upgrades require validation; (2) no single model-agent combination dominates: best performances range from 57% to 100% depending on domain. These examples illustrate how SOP-Bench enables isolating and studying specific dimensions of agent performance without costly production experiments. Our goal is not to rank model capabilities or build optimal agents, but to provide a rigorous evaluation framework that enables the researchers and practitioners to systematically investigate agent design choices, model selection, and deployment strategies. We release the benchmark at https://github.com/amazon-science/sop-bench.

cs.AI↗

arXiv2026

BeDiscovER: The Benchmark of Discourse Understanding in the Era of Reasoning Language Models

We introduce BeDiscovER (Benchmark of Discourse Understanding in the Era of Reasoning Language Models), an up-to-date, comprehensive suite for evaluating the discourse-level knowledge of modern LLMs. BeDiscovER compiles 5 publicly available discourse tasks across discourse lexicon, (multi-)sentential, and documental levels, with in total 52 individual datasets. It covers both extensively studied tasks such as discourse parsing and temporal relation extraction, as well as some novel challenges such as discourse particle disambiguation (e.g., ``just''), and also aggregates a shared task on Discourse Relation Parsing and Treebanking for multilingual and multi-framework discourse relation classification. We evaluate open-source LLMs: Qwen3 series, DeepSeek-R1, and frontier model such as GPT-5-mini on BeDiscovER, and find that state-of-the-art models exhibit strong performance in arithmetic aspect of temporal reasoning, but they struggle with full document reasoning and some subtle semantic and discourse phenomena, such as rhetorical relation recognition.

cs.CL↗

arXiv2025

AgentAda: Skill-Adaptive Data Analytics for Tailored Insight Discovery

We introduce AgentAda, the first LLM-powered analytics agent that can learn and use new analytics skills to extract more specialized insights. Unlike existing methods that require users to manually decide which data analytics method to apply, AgentAda automatically identifies the skill needed from a library of analytical skills to perform the analysis. This also allows AgentAda to use skills that existing LLMs cannot perform out of the box. The library covers a range of methods, including clustering, predictive modeling, and NLP techniques like BERT, which allow AgentAda to handle complex analytics tasks based on what the user needs. AgentAda's dataset-to-insight extraction strategy consists of three key steps: (I) a question generator to generate queries relevant to the user's goal and persona, (II) a hybrid Retrieval-Augmented Generation (RAG)-based skill matcher to choose the best data analytics skill from the skill library, and (III) a code generator that produces executable code based on the retrieved skill's documentation to extract key patterns. We also introduce KaggleBench, a benchmark of curated notebooks across diverse domains, to evaluate AgentAda's performance. We conducted a human evaluation demonstrating that AgentAda provides more insightful analytics than existing tools, with 48.78% of evaluators preferring its analyses, compared to 27.67% for the unskilled agent. We also propose a novel LLM-as-a-judge approach that we show is aligned with human evaluation as a way to automate insight quality evaluation at larger scale.

cs.CL↗

arXiv2025

CEMTM: Contextual Embedding-based Multimodal Topic Modeling

We introduce CEMTM, a context-enhanced multimodal topic model designed to infer coherent and interpretable topic structures from both short and long documents containing text and images. CEMTM builds on fine-tuned large vision language models (LVLMs) to obtain contextualized embeddings, and employs a distributional attention mechanism to weight token-level contributions to topic inference. A reconstruction objective aligns topic-based representations with the document embedding, encouraging semantic consistency across modalities. Unlike existing approaches, CEMTM can process multiple images per document without repeated encoding and maintains interpretability through explicit word-topic and document-topic distributions. Extensive experiments on six multimodal benchmarks show that CEMTM consistently outperforms unimodal and multimodal baselines, achieving a remarkable average LLM score of 2.61. Further analysis shows its effectiveness in downstream few-shot retrieval and its ability to capture visually grounded semantics in complex domains such as scientific articles.

cs.CL↗