Peer Revenue Past citations are no guarantee of future returns

Papers

Fresh from arXiv, typeset properly. Highlighted words trade on the market.

Computer Science Artificial Intelligence

cs.AI Watch Artificial Intelligence

Wednesday 17 June

Artificial Intelligence · Cross-listed from cs.CL

Multi-Modal Generative Fuzzy System: Fuzzy Inference Guided Large Model Interactive Question Answering Framework

Hailong Yang, Jianqi Wang, Guanjin Wang et al. (+1 more)

In Multimodal Question Answering (MQA), models are required to jointly encode and integrate heterogeneous information from multiple modalities, including text, images, and speech, to perform complex semantic reasoning and decision making. Despite recent advances, existing approaches, including traditional deep learning models and Large Models (LMs) or prompt-based frameworks, continue to face several critical challenges. First, modality bias arises from discrepancies in feature distributions across different modalities, which limits effective cross modal collaborative understanding. Second, many questions require knowledge drawn from multiple domains, introducing significant uncertainty. Third, current methods often rely on shallow semantic matching, resulting in limited reasoning depth an reduced interpretability. To address these issues, inspired by the traditional fuzzy system (FS) framework, we propose a fuzzy-inference-guided multimodal generative architecture termed the Multi-Modal Generative Fuzzy System (MMGFS). The main contributions of MMGFS are two folds. First, it alleviates modality bias through a multimodal collaborative rumination mechanism. Second, it introduces fuzzy rules and a multi-hop inference mechanism to support cross-domain knowledge fusion and hierarchical reasoning, thereby strengthening uncertainty modelling and deepening semantic understanding. We conduct comprehensive evaluations on open-domain question answering datasets, including MultimodalQA and WebQA, as well as domain-specific benchmarks, including BioMol-VQA and EHRxQA. Experimental results demonstrate that MMGFS consistently outperforms existing methods across multiple datasets. It effectively mitigates modality bias and question uncertainty while achieving superior performance in answer accuracy, consistency, and generalization.
Read the full abstract

All authors Hailong Yang, Jianqi Wang, Guanjin Wang, Zhaohong Deng

Artificial Intelligence · Cross-listed from cs.DB

Can Large Language Models Reason about Event-Time Stream-Processing Semantics?

Zhuoxi Wang, Shibo Zheng and Haoyu Zhang

Streaming systems increasingly hand work to large language models (LLMs): writing pipelines, triaging alerts, reading logs. All of it assumes the model knows how event-time stream processing behaves, and we test that assumption directly. StreamReason-Bench asks a model to stand in for an event-time stream processor. Given a windowed query and a stream of out-of-order events, it reports which windows fire, with their aggregates, and which events are dropped as late. The answer key comes from a small reference implementation of Dataflow-model semantics, so we can grade exactly, and with a partial-credit row-F1, without running an engine. On 600 generated items covering tumbling, hopping, session, and processing-time windows, the models do poorly on event time. When told to answer directly, no model that actually follows the instruction clears 34% exact match; chain-of-thought (CoT) roughly doubles that for several of them (GPT-4o goes from 0.34 to 0.48), and only one frontier model that reasons by default comes near solving the set (0.85). A processing-time control, with no watermarks and nothing late, is almost solved by every capable model. The gap points to event-time and late-data handling, not windowing or arithmetic, as the hard part. Sorting errors by window type tells the same story: late-data mistakes dominate the event-time windows and vanish on the control, while session windows mostly fail on where the session boundaries fall.
Read the full abstract

Artificial Intelligence · Cross-listed from cs.HC

EmoPatient: An Emotion-Directed Patient Simulator for Realistic Palliative Care Communication Training

Yining Wu, Tianshu Du, Jinrui Fang et al. (+3 more)

Effective communication during palliative care discussions is a critical clinical skill, yet training clinicians to manage complex patient emotions remains challenging. Large language model (LLM)-based patient simulators provide a scalable approach for communication training, but most existing systems treat patient emotion as static and fail to capture the dynamic emotional shifts observed in clinical interactions. We present EmoPatient, an emotion-directed patient simulator designed to generate evolving emotional responses during palliative care discussions. The system introduces an Emotion Director agent that estimates the patient's emotional state and generates turn-level control signals for emotional intensity, regulatory stability, and interactional guidance. We evaluate EmoPatient through controlled multi-turn physician-patient dialogue simulations and compare it with baseline simulators. Results show improvements across four theory-informed emotional realism metrics and robustness across conversational personality variants, suggesting that modeling emotional dynamics can improve the realism of LLM-based patient simulators for palliative care communication training.
Read the full abstract

All authors Yining Wu, Tianshu Du, Jinrui Fang, Chi Zhang, Sonal Admane, Ying Ding

Artificial Intelligence · Cross-listed from cs.HC

How to Ask the AI: A User Perspective Survey for Large Language Model Prompting

Yiqun Zhang, Yunfan Zhang, Mingjie Zhao et al. (+2 more)

AI tools like ChatGPT and DeepSeek, powered by Large Language Models (LLMs), allow users to obtain instant and effective content responses simply by typing requests, such as “plan a three-day Vienna trip”, “solve the attached mathematical problem”, “draft an email to inquire review progress”, etc., which are also known as LLM prompts. Crafting clear and well-structured prompts leads to more appropriate LLM feedback, which effectively bridges human-LLM interaction. Although prompting appears accessible to non-expert users, precisely organizing effective prompts is a highly systematic and skillful process, presenting potential challenges even for experienced users. This survey explores the principles, taxonomy, and organization of prompts from a user-centered perspective. Differing from the existing surveys that primarily focus on technical principles and application scenarios of LLMs, this paper provides actionable guidelines for formulating effective LLM prompts across diverse real-world tasks and specifically contributes by: 1) developing an intuitive evaluation strategy for prompt efficacy, 2) providing prompting workflow demonstrations on representative applications, and 3) maintaining a dynamically updated open-source project to ensure the core takeaways remain up-to-date. These measures lower the threshold for users to correctly understand and craft prompts that align with evolving application scenarios. This work will be maintained as a living GitHub project \href{https://github.com/Yunfan-Zhang/TAI_Guideline-Table}{\textcolor{blue}{here}}.
Read the full abstract

All authors Yiqun Zhang, Yunfan Zhang, Mingjie Zhao, Sen Feng, Yiu-ming Cheung

Artificial Intelligence · Cross-listed from cs.CY

Archetypes or ability? Clustering for modelling student mathematical competence

Benjamin Mawdsley, Tom Quilter, Richard Turner et al. (+2 more)

Personalised learning systems often assume that mathematical ability is combined of discrete abilities, acquired sequentially and dependent upon first acquiring foundational abilities, and students often report different strengths. In this work, we explore the validity of these assumptions by applying clustering methods to a large dataset of 119,034 students, spanning 13 national-level exams sat in the United Kingdom and collected by the platform. Classifying question results as pass or fail, we use a Bernoulli Mixture Model to search for latent populations which would be indicative of discrete skill-sets. We find that few distinct clusters are present in the data and that the dominant factor is overall student ability, which is further supported by the high degree of linear correlation between the probability distributions of the resulting clusters. Our best performing model achieves an accuracy of 78 percent, competitive with more complicated models in the literature whilst being more explainable. Comparing this models performance with logistic regression baselines and with k-nearest neighbours, we find a small improvement when using performance on each individual question as features. This suggests that whilst overall ability level is the dominant factor for predicting performance, small further personalisation improvements can be made by tailoring to a students exact strengths, but that students do not appear to develop strongly differing ability across topics. Our work offers a national scale test of machine learning in education and offers a new benchmark for the field, demonstrating how explainable models can reach competitive performance
Read the full abstract

All authors Benjamin Mawdsley, Tom Quilter, Richard Turner, Sarah Jackson, Paul Edwards

Artificial Intelligence

HOBA: Hierarchical On-Policy Bidding Agents for Adaptive Online Advertising

Ji Wu, Yunshan Peng, Wentao Bai et al. (+5 more)

Online advertising bidding systems typically deploy multiple offline-trained expert models (e.g., PID controllers, model predictive control, offline RL policies) but face two critical limitations: lack of online adaptability to non-stationary auction markets, and reliance on costly manual tuning of hyperparameters such as bid bounds and budget pacing constraints. We propose HOBA (Hierarchical On-policy Bidding Agents), a hierarchical reinforcement learning framework that decouples strategic reasoning, model selection, and bid execution across three time scales. At the high level, a large language model infers hyperparameters from contextual signals through a Think-Act-Observe-Reflect loop with historical experience retrieval. At the mid level, a SARSA agent dynamically selects among expert models, incorporating causal adjustment to eliminate selection bias. At the low level, a dynamic expert pool (PID, MPC, IQL, Decision Transformer) executes bids under high-level constraints. This design confines online learning to discrete expert selection rather than continuous bid optimization, significantly reducing exploration risk while maintaining adaptability. Experiments on the AuctionNet benchmark and a large-scale A/B test demonstrate consistent improvements over state-of-the-art baselines. In a large-scale online deployment, HOBA delivered substantial business value, achieving a +3.6% increase in target cost, proving the effectiveness of our hierarchical multi-agent bidding paradigm.
Read the full abstract

All authors Ji Wu, Yunshan Peng, Wentao Bai, Yunke Bai, Wenzheng Shu, Jinan Pang, Yanxiang Zeng, Xialong Liu

Artificial Intelligence

TokenMem: Faithful Knowledge Injection for Frozen LLMs

Chengzhang Yu, Chenyang Zheng, Zening Lu et al. (+5 more)

Retrieval-augmented generation (RAG) enhances large language models (LLMs) with external knowledge, but suffers from knowledge conflicts: when retrieved information contradicts parametric memory, the shared self-attention pathway produces unpredictable outputs. We present TokenMem, a lightweight memory system that injects knowledge into frozen LLMs through a dedicated cross-attention channel, bypassing competition with parametric memory in the residual stream. TokenMem trains only a thin gating adapter ($\sim$3-7M parameters) via a two-phase curriculum: first learning general knowledge utilization, then strengthening faithful compliance under counterfactual knowledge. In controlled experiments on five models spanning three families (Qwen3-4B/8B/14B, LLaMA-3.1-8B, OLMo-3-7B), TokenMem achieves 69-70% Knowledge Compliance (KC) on counterfactual benchmarks, compared to 20-52% for vanilla RAG, a gap of up to 49 percentage points. Ablation studies show that the two-phase curriculum is critical: removing Phase 2 collapses KC to near-zero. Mechanistic analysis reveals that the gate adapter learns a conflict-aware, layer-specific injection strategy without explicit supervision.
Read the full abstract

All authors Chengzhang Yu, Chenyang Zheng, Zening Lu, Yingru He, Yutong Huang, Yiming Zhang, Yue Xu, Zhanpeng Jin

Artificial Intelligence

CHS-SQL: A Text-to-SQL approach based on Confidence-Guided Heuristic Search Schema Linking process

Minghao Yang and Yanjun Xu

Recently, there have been several works in the Text-to-SQL domain that utilize Small Language Models (SLMs) for training. These approaches achieve performance close to that of large models in generating SQL, using only the computational power of a single NVIDIA RTX 4090 GPU, while also ensuring data security. Most existing methods filter out redundant tables and columns during Schema Linking to improve Text-to-SQL accuracy. However, they do not consider the precision-recall trade-off when selecting the candidate schema subset. Our research found that both the precision and recall of Schema Linking directly affect the final SQL accuracy. Therefore, we propose a novel framework for efficiently fine-tuning SLMs on Text-to-SQL tasks, CHS-SQL, that not only balances precision and recall but also improves overall performance on Text-to-SQL tasks. Its main innovation lies in the Schema Linking phase, where a heuristic search combined with model internal confidence is employed to achieve an optimal precision-recall trade-off. This elaborated mechanism maximizes the precision of relevant schema candidates for the generated SQL queries while suppressing irrelevant noise. The same strategy is further applied during SQL generation to refine candidate queries while helping the SLM to avoid trapping in a local optimum. Our method achieves state-of-the-art (SOTA) results on Text-to-SQL tasks via SLMs.
Read the full abstract

Artificial Intelligence · Cross-listed from cs.CL

Learning When to Reason for Text-to-SQL via SFT and DPO

Soohyuk Jang, Jiheum Yeom, Nohil Park et al. (+4 more)

Recent Text-to-SQL methods rely heavily on reasoning-centric paradigms such as Chain-of-Thought (CoT), achieving substantial gains on complex benchmarks at the cost of high inference-time overhead. However, a large fraction of real-world queries are simple lookups or aggregations that can be resolved without multi-step deduction, making forced reasoning wasteful. Thus, we propose AutoThinkSQL, a framework that integrates an auto-thinking mechanism into both Supervised Fine-Tuning (SFT) and Direct Preference Optimization (DPO) on Text-to-SQL. Our approach enables the model to dynamically bypass reasoning for simple queries while invoking deep CoT for complex queries. On Qwen3-Coder-30B-A3B, our method achieves consistent gains compared to the best counterpart baseline on both Spider and BIRD benchmarks while simultaneously reducing average output tokens by 24.6% and 18.3%, and average latency by 17.1% and 11.5% compared to CoT-only generation. Further analysis indicates that the model learns to align its reasoning decisions with query difficulty.
Read the full abstract

All authors Soohyuk Jang, Jiheum Yeom, Nohil Park, Sang Hun Kim, Yoonyoung Choi, Kiwook Bae, Sungroh Yoon

Artificial Intelligence

Opti-Q: A Constraint-Based Optimization Framework for Multi-LLM Question Planning

Aamir Hamid, Bharg Barot, Satvik Racharla et al. (+3 more)

While large language models (LLMs) enable strong question answering (QA), budgeted deployment is complicated by nondeterminism and heterogeneous resource profiles (cost, latency, and energy). We present OPTI-Q, a database-inspired, cost-based optimizer that implements a plan-before-execute paradigm for multi-LLM orchestration. OPTI-Q models LLM invocations as physical operators in an execution DAG and, for each question, searches for plans that optimize answer quality (QoA) while trading off financial cost, latency, and energy under user-specified resource constraints. Plans can include sequential operators that pass intermediate answers as context and parallel/blend operators that run models concurrently and merge their outputs. To search this space without executing each candidate plan, OPTI-Q uses PERFDB, a statistics catalog populated and refreshed from benchmarks and execution traces, to estimate the QoA and resource costs of both individual operators and composed subplans. Using these estimates, OPTI-Q performs Pareto-frontier search and selects a final plan based on user preferences. On MMLU-Pro and SimpleQA under user-specified budgets, OPTI-Q improves average QoA by ~58% and ~41% over baselines at comparable cost, demonstrating that database-style planning yields better quality-resource trade-offs for multi-LLM QA.
Read the full abstract

All authors Aamir Hamid, Bharg Barot, Satvik Racharla, Tim Finin, Primal Pappachan, Roberto Yus

Artificial Intelligence

Euclean: Automated Geometry Problem Formalization with Unified Verification in Lean

Linbin Tang, Jingyan You, Zilin Kang et al. (+8 more)

Recent formal reasoning systems have reached IMO-level performance, yet they leave a fragmented landscape: algebra and number theory are handled in Lean, while geometry still relies on domain-specific languages with limited formal guarantees. This split increases the trusted computing base and hinders unified model development. Existing geometry-in-Lean efforts (LeanEuclid, LeanGeo) introduce custom axiom systems incompatible with standard Mathlib, and their small scale ($<$ 1,100 problems) limits large-scale training. Native Mathlib autoformalization of geometry, however, poses distinct challenges: implicit diagrammatic assumptions (e.g., topological configuration and non-degeneracy) must be made explicit rather than deferred to external solvers, and models must adapt to Mathlib's small, rapidly evolving geometry infrastructure. We present Euclean, a four-stage framework - constraint explication, configuration anchoring, formalization mapping, and iterative repair - for automatically formalizing geometry in native Mathlib. We construct OMNI-Geometry (768 competition problems) and Numina-Geometry (177,597 problems), the largest geometry formalization dataset in Lean. Human evaluation shows 48.89% TOP1 and 73.33% TOP5 accuracy. Training Goedel v2 on our formalizations improves proof success from 13.6% to 15.1%, validating dataset quality for unified neural theorem proving. Code and datasets: https://github.com/tlb-22/Euclean.
Read the full abstract

All authors Linbin Tang, Jingyan You, Zilin Kang, Hanzhang Liu, Sophia Zhang, Zenan Li, Chenrui Cao, Liangcheng Song, Jiaao Wu, Xian Zhang, Fan Yang

Artificial Intelligence · Cross-listed from cs.GT

A Calculus of Discernment: Decision-Relevant Insight, Sequence Value, and Forgetting as Higher-Order Learning

Suyash Mishra

In a world of generative AI, candidate insights are abundant; what is scarce is the capacity to discern which matter, to act on them in the right amount and order, and to forget the rest so the system can adapt. We argue these scarcities are governed by one object and build a framework around it. We define an insight strictly as a lever with an identified, measurable effect on an objective, and rank candidates by decision-relevance via the expected value of information rather than novelty. We show action carries an order, not only a size: under realistic belief dynamics, content "touches" are non-commuting operators, so a fixed plan delivered in different orders yields different outcomes, defining a sequence premium. We observe that the value of any lever is a shadow price, unifying pharmaceutical marketing, equity selection, and manufacturing as one leverage-discovery problem. Most speculatively, we propose APOHA, a theory in which forgetting is not the disposal of knowledge but the operator by which value is learned: the value of a retained item is the counterfactual cost of forgetting it, a learning system is the residue of maximal forgetting subject to preserved value, and higher-order value is the structure that survives repeated forgetting (a renormalisation-relevant invariant), with consolidation as its conjugate. We state the central open problem (a non-trivial attractor with a spectral gap) and test the forgetting theory: operationalising APOHA as an agent on a non-stationary obesity-treatment decision world over 30 seeds, adaptive forgetting cut cumulative decision-regret by 24-32% against never-forget and a fixed half-life, kept a ~6x smaller, cleaner memory, and converged stably; notably, blind forgetting was worse than never forgetting, so the benefit is specific to value-aware forgetting. A multi-disciplinary critique stress-tests the whole.
Read the full abstract

Artificial Intelligence · Cross-listed from cs.LG

Fully-sensorized smart-eyewear platform for on-device Machine Learning

Andrea Giudici, Christian Veronesi, Pietro Bartoli et al. (+5 more)

This paper presents ARGO, a smart eyewear platform designed to bridge ergonomic comfort, high computational throughput, and energy efficiency. Unlike cloud-dependent solutions, ARGO leverages the STM32N6 microcontroller and its integrated Neural Processing Unit (NPU) to enable on-device machine learning, minimizing latency and preserving user privacy through local data processing. The primary contribution lies in the holistic co-design of hardware, firmware, and artificial intelligence, centered on the deployment of an optimized YOLOv11 model for real-time urban obstacle recognition. To ensure compatibility with the target NPU, we introduce Head-wise Parallel Attention (HPA), an architectural refinement that enables efficient accelerator execution while preserving the original computational logic. The model is trained on the Walking On The Road (WOTR) dataset, and the final deployed configuration achieves an mAP50-95 of 24 under strict memory constraints, with a memory footprint of only 2.483 MB. The platform integrates a multimodal sensor suite, RGB cameras, Time-of-Flight sensors, microphones, and ambient sensors, and delivers 10 FPS at a continuous autonomy of ~113 minutes on a 200 mAh battery. These results demonstrate the feasibility of a high-performance, privacy-preserving, and socially acceptable assistive device, and highlight how competitive edge AI solutions increasingly demand tightly integrated, multidisciplinary co-design approaches.
Read the full abstract

All authors Andrea Giudici, Christian Veronesi, Pietro Bartoli, Mario Caliò, Aurelio Teliti, Giacomo Gervasoni, Diana Trojaniello, Franco Zappa

Artificial Intelligence

Evidence-Aware MapReduce for Forkable Compute

Yossi Eliaz

Snapshot-backed sandboxes make branching cheap while leaving evidence dependence unchanged. Branches can reuse a model, prompt, repository, tests, observations, or execution ancestor, so counting outputs can amplify one repeated error into high-confidence consensus. We introduce an evidence-aware reduction contract: each worker reports an estimate, estimated information, evidence identifiers, fork lineage, and execution metadata. For independent workers estimating one common parameter, we use standard inverse-information pooling in its Gaussian/Wald form. The fixed-dimensional numeric summary can merge in any tree order; evidence IDs and lineage follow separate rules. The residual $\Delta$ measures disagreement, becomes Cochran's $Q$ in the scalar inverse-variance case, and appears in the product integral. A reference implementation validates serialized records, rejects repeated nonempty evidence identifiers, carries evidence and lineage through tree reduction, and uses Cholesky-based numerical linear algebra. Unit tests and seeded synthetic checks exercise the algebra, unequal information, and forged precision; one four-worker named-snapshot trace exercises the end-to-end path. Platform logs document the exercised execution paths. A central open systems challenge is to turn evidence identity and fork lineage into a dependence model for correlated and adaptively selected AI branches.
Read the full abstract

Artificial Intelligence · Cross-listed from stat.AP

Estimating the Effect of Timing on Coupon Effectiveness

Deddy Jobson

The coupon incentive is one of the most common tools marketers use to court users to engage with a business at various stages of the customer life cycle. A variety of factors can affect the effectiveness of a coupon incentive on users, timing being one of them. We hypothesize that coupons can be more effective when delivered at critical times in the customer journey, right when a user is engaging with the platform. Verifying such a hypothesis would typically require real time event-triggered coupon distribution software that may be too expensive to implement. In this paper, we propose a framework in which we apply causal inference on "natural randomized control trial experiments" to measure the effectiveness of sending coupons at the right time to users without requiring a dedicated AB test. We demonstrate the usefulness of our framework in the case of a user onboarding coupon campaign held in our company and show how the results can lead to correct data-driven decisions for the business. Furthermore, in order to test the generalizability of our framework, and to make our research more reproducible, we apply our framework on a user retention campaign with a publicly available dataset.
Read the full abstract

Artificial Intelligence · Cross-listed from cs.CY

ELEVATE: Designing Human-Centered GenAI Virtual Tutors for Scalable and Inclusive Education

Lorenzo Stacchio, Michele Giordano, Daniele Berardini et al. (+2 more)

The advent of Generative Artificial Intelligence (GenAI), and in particular Large Language Models (LLMs), is reshaping educational practice, while intensifying ethical debate about its adoption. To date, the dominant paradigm remains cloud-based and text-only chatbot: a centralized service that offers limited pedagogical control, weak transparency over knowledge sources, and non-trivial risks for privacy and regulatory compliance. This model also presumes continuous connectivity and recurring API costs, creating structural barriers for many institutions, reinforcing existing digital divides. At the same time, educational interaction with LLM can benefit from multimodal cues and embodied presence, requiring interfaces that move beyond text-only tutoring. In this work, we propose ELEVATE (Efficient LLM Education with Virtual Avatar Teaching Engine), a framework to develop efficient GenAI-driven avatar tutors governed by epistemic infrastructures. ELEVATE integrates LLM-driven dialogue with embodied 3D avatars for multimodal interaction and adopts a local-first execution model enabling deployment on consumer-grade hardware. The framework formalizes a three-stratum design that separates (i) a student-facing virtual avatar interaction layer, (ii) a local GenAI execution and multimodal synthesis core, and (iii) a teacher-facing governance layer. We implemented and evaluated a working prototype deployed in a real-world educational curriculum. The system runs on standard PCs and smartphones, and we provide system-level performance evidence to show responsive interaction under realistic hardware constraints. Finally, we discuss sociotechnical and pedagogical implications for responsible adoption, positioning ELEVATE as a scalable pathway for privacy-preserving and inclusive GenAI tutoring across heterogeneous school environments.
Read the full abstract

All authors Lorenzo Stacchio, Michele Giordano, Daniele Berardini, Primo Zingaretti, Emanuele Frontoni

Artificial Intelligence

Recursive Self-Evolving Agents via Held-Out Selection

Michael Nguyen, Quoc Nguyen and Paul Vuong

LLM agents are increasingly improved without weight updates by evolving a natural-language artifact, such as reflections, workflows, playbooks, cheatsheets, or optimized prompts, that conditions a frozen policy. Such methods are typically reported as wins on the single benchmark where they help. We study them apples-to-apples and surface a sharper picture. We introduce RSEA, a Recursive Self-Evolving Agent that carries a compact three-layer natural-language state: an imperative strategy, reusable skills, and a procedural playbook. Across generations, RSEA rewrites all three layers from its own trajectories and commits a candidate only if it does not regress on a disjoint held-out split, using a strict keep-better gate. Across four diverse benchmarks, ALFWorld, GAIA, (\tau)-bench, and WebShop, and six faithful baselines, ReAct, Reflexion, GEPA, AWM, ACE, and Dynamic Cheatsheet, all evaluated on one shared local backbone, we find three main results. First, no artifact universally wins. RSEA is the strongest single-pass method on ALFWorld, reaching 69.3% compared with 64.6% for ReAct (McNemar (p=0.015)), and reaches 79.4% with retry, the best overall result. However, concrete-workflow induction, represented by AWM, is best on the strong-backbone tool-use tasks. Second, unguarded context evolution is high-variance and unsafe. Dynamic Cheatsheet, which curates context online without a held-out gate, is near-best on ALFWorld at 70.7%, yet collapses on WebShop, with a score of 0.14 compared with 0.43 for ReAct. Third, RSEA's strict held-out selection is what makes recursive self-evolution monotone-safe: it never significantly underperforms the base agent on any benchmark and falls back to vanilla ReAct when evolved context would hurt.
Read the full abstract

Artificial Intelligence · Cross-listed from cs.NE

Model Merging to Evolution: Parameter Space Exploration for Expert Models

Chao Wang, Yuchen Guo, Zheng Tan et al. (+4 more)

Model merging integrates the capabilities of multiple expert models to create strong models for multiple tasks without additional training, thereby reducing computational resource requirements. However, existing methods operate within the convex combination space of expert models, failing to explore high-performance regions outside this space. This paper proposes the MERGEvolve framework, which unifies model merging and evolution within an evolution strategy by treating the merged model as the initialization for evolutionary exploration of the parameter space. During the merging phase, expert models act as deterministic sources to build a strong initial point. The evolution phase then explores the parameter space using random noise. Theoretical analysis shows that MERGEvolve explores regions outside the convex combination space. Extensive experiments on single-task and multi-task benchmarks demonstrate that MERGEvolve consistently achieves performance competitive with advanced model merging baselines. Ablation studies confirm that a high-quality initial point is critical for efficient exploration of the parameter space.
Read the full abstract

All authors Chao Wang, Yuchen Guo, Zheng Tan, Guanchun Wang, Yanbiao Ma, Qiqi Duan, Peng Wu

Artificial Intelligence

Self-Improvement Can Self-Regress: The Rise-and-Collapse Failure Mode of LLM Self-Training

Jianzhe Lin

Self-improvement can self-regress. In REINFORCE post-training for code, a model can quickly improve on its optimized metric and then collapse within the same training campaign. We study this in a controlled multi-seed testbed using Qwen-2.5-3B and Qwen-2.5-7B, trained on competitive-programming tasks with binary CodeGrader reward across 10 sequential 20-step campaigns. Across campaigns, pass@1 shows a robust rise-then-collapse pattern: it peaks within tens of gradient steps and then falls back, sometimes to near zero. This is not cross-task catastrophic forgetting, but within-task policy over-optimization on a fixed distribution; KL- and EWC-style constraints do not prevent it. We ask where the control loop should sit. We compare three levels: CARE, a between-campaign memory mechanism with a capability posterior, transfer gate, and regression-aware belief revision; ES, a within-campaign early-stop rule that rolls forward the peak checkpoint and sets the next budget to peak_step+3; and GRPO, which changes the RL update using group-relative reward normalization. The answer is regime-dependent. On Qwen-2.5-3B, where naive REINFORCE is fragile, CARE v2 nearly doubles end-of-chain pass@1 from 4.9% to 9.5%, with paired bootstrap 95% CI [+0.4,+8.9] and gains in 4/5 seeds. On Qwen-2.5-7B, CARE reaches parity with naive REINFORCE, 13.8% vs. 11.8%, while ES reaches 22.2% [14.1,28.0]. Out-of-the-box GRPO reaches 20.7% [15.7,25.1], nearly matching REINFORCE+ES. GRPO raises the floor but does not remove the cliff. Its 7B gain mainly comes from better between-campaign carryover, while the within-campaign peak-to-end gap remains about 17 points under both REINFORCE and GRPO. GRPO+ES gives mixed evidence: 2/3 seeds improve, but one final cliff lowers the mean to 17.0% [0.0,28.1]. A Gemma-3-4B pilot shows the same signature, suggesting the phenomenon is not limited to Qwen.
Read the full abstract

Artificial Intelligence

Repeated post-training is not Self-improving: Diagnosing Scientific Amnesia in Continual DPO Pipelines

Jianzhe Lin, Fei Wang, Xiaolin Li et al. (+2 more)

Industrial LLM teams often ship behavior updates by repeatedly DPO-training a base model on sequences of related preference-data campaigns. The dominant failure mode in this regime is not always classical catastrophic forgetting: a pipeline may preserve previously learned behaviors while still failing to accumulate reusable methodological knowledge about how to train the next campaign. We call this failure mode scientific amnesia. This paper turns that practitioner intuition into a measurable industrial problem. We contribute: (i) a diagnostic suite for amnesia, (ii) a Program-based pipeline that chains FSDP-sharded DPO checkpoints across Qwen2.5-7B-Instruct runs, (iii) a 30-campaign HumanEval subdomain benchmark, and (iv) a comparative diagnostic study of five strategy proposers: random memory, rule-based scheduling, retrieval-only memory, warm-start Bayesian optimization, and MSCL, a meta-scientific memory and reasoner candidate. Across a single-seed 5-condition * 3-step real-LM chain, 4 of 5 candidates degrade in step-level peak pass@1, including MSCL; only the deliberately conservative rule-based schedule improves. Follow-up pilots qualify rather than overturn this finding: in a heterogeneous chain, MSCL is the only completed candidate that improves, whereas in a small multi-seed homogeneous sweep, retrieval-only has the best mean Delta and no pairwise candidate gap is statistically distinguishable. The contribution is therefore diagnostic, not a claim that MSCL solves the problem: scientific amnesia is observable in a production-like continual-DPO pipeline, and conclusions about interventions depend sharply on chain regime, evaluator design, and seed coverage.
Read the full abstract

All authors Jianzhe Lin, Fei Wang, Xiaolin Li, Rajeshkumar Golani, Jubin Chheda

Artificial Intelligence · Cross-listed from cs.LG

Massive Activations Are Architecturally Robust: A Controlled Scratch/Commitment Residual Stream Test

Maruthi Vemula

Trained transformers reliably develop massive activations, a small number of hidden dimensions whose magnitude is far above the median and which concentrate on the sequence-start token. Whether these outliers are a removable artifact of the residual stream's overloaded read and write role, or instead a functional necessity, is actively debated. We test the artifact hypothesis directly, with an architectural intervention. Our architecture, Ledger Residuals, splits the residual stream into a mutable scratch stream (Deliberation) that intermediate computation may freely overwrite and a protected, decode-only accumulator (Commitment) that holds the representation the model reads out. If massive activations exist only because one stream is forced to be both scratchpad and answer, then a dedicated answer channel should remove the need for them. We find that it does not. In matched-loss language models at the 160M and 290M scales, the model rebuilds the canonical fixed-dimension, start-token outlier inside the protected channel. The rebuilt feature is smaller in magnitude than in a standard transformer but more sharply concentrated on the start token, and a stronger sparsity penalty makes it more persistent and more concentrated still, rather than removing it. Massive activations therefore look architecturally robust: they re-emerge in whichever representation the model decodes from, which is what we would expect if they are functional rather than incidental. We release our architecture and measurement code.
Read the full abstract

Artificial Intelligence · Cross-listed from cs.RO

A Digital Twin Framework for Traffic-Aware UAV Pavement Monitoring in Open-Traffic Conditions

Yamil Uchani, Grace Luna, Edwin Salcedo et al. (+1 more)

UAV-based pavement inspection can reduce the cost and risk of road-surface monitoring, but real-world deployment remains difficult when traffic, pedestrians, and temporary occlusions affect defect visibility. This paper presents a Unity-based digital twin framework for traffic-aware UAV pavement monitoring in open-traffic conditions. The proposed environment integrates procedurally generated road defects, dynamic traffic agents, autonomous UAV navigation, and a multitask YOLOv8n perception module for detecting road defects, pedestrians, and vehicles while classifying road-defect subtypes. After synthetic-domain fine-tuning, the perception model achieved 0.959 mAP@0.5 and 0.940 macro F1-score on a held-out synthetic test set generated from the simulator. The digital twin was then used to evaluate hover-and-recheck, micro-repositioning, and skip-and-revisit recovery strategies across different traffic densities and flight altitudes. Results show that flight altitude strongly affects inspection coverage, while recovery strategies introduce different trade-offs between coverage, mission duration, energy consumption, and revisit behaviour. These findings demonstrate that digital twins can support the development and evaluation of traffic-aware UAV inspection strategies before real-world deployment. The full implementation and trained models are available at https://github.com/EdwinTSalcedo/RDMO-DigitalTwin.
Read the full abstract

All authors Yamil Uchani, Grace Luna, Edwin Salcedo, Mauricio Figueroa

Artificial Intelligence · Cross-listed from cs.CV

An approach with Visual and Tabular Mamba to multimodal medical data using Mixed Fusion

Matheus B. Rocha, Gustavo B. Dettogni and Renato A. Krohling

This article presents a complementary approach for integrating multimodal medical data in cancer classification, based on state space models represented by the Mamba architecture. To this end, a mixed multimodal fusion architecture, called Mixed Fusion, was employed and developed to enhance the interpretability of the decision-making process. The proposed approach explores two variants of Mamba: one dedicated to visual processing, responsible for classifying the lesion image and generating probabilities associated with the target classes, and another focused on tabular processing, which uses these probabilities together with clinical and/or sociodemographic data to produce the final diagnosis. The experiments were conducted on two medical datasets: PAD-UFES-20, composed of clinical images and information associated with skin lesions, and NDB-UFES, consisting of histopathological images and sociodemographic data related to oral cancer. The results indicate slightly lower performance in balanced accuracy, compared with Transformer-based approaches, on PAD-UFES-20, and superior performance on NDB-UFES. Additionally, substantial gains were observed in the recall metric. Furthermore, the adoption of the Mixed Fusion architecture enables the application of the Shapley Additive Explanations (SHAP) method, increasing the interpretability of the results. These findings indicate that Mamba-based models constitute a suitable alternative for multimodal classification in medical data, especially in scenarios in which sensitivity is a relevant requirement.
Read the full abstract

Artificial Intelligence

Repeated Shared Access Enables Grokking, but Edit Propagation Depends on an Addressable Memory

Yanan Niu

We study factual edit propagation in a controlled synthetic knowledge-graph QA setting using a 2x2 grid that crosses loop recurrence with shared-memory access: a dense transformer (Dense), a looped transformer (Loop), a dense backbone with shared memory (Dense+Mem), and a looped backbone with shared memory (loop-memory coupling, LMC). The two factors dissociate. For learning, both routes to repeated shared access – looped recomputation and repeated memory rereading – cross the out-of-distribution (OOD) grokking barrier that Dense fails, so repeated shared access is the behavioral regularity, not a specific architecture. For editing, the substrates split along a different axis: applying a single localized factual edit (conditioned on direct success) and measuring 2-hop propagation on a shared pre-edit-correct set, the edit propagates strongly in both memory-bearing cells (LMC 0.78-0.92, Dense+Mem 0.71-0.96) and only weakly in the memory-free ones (Loop 0.04-0.30, Dense 0.00-0.03). The split is along the memory axis, not the loop axis: every memory-bearing seed exceeds every memory-free seed, with no detectable difference between the two memory cells. Crucially Dense+Mem has no recurrence, so the propagating ingredient is an addressable site that an edit can write to and later computation rereads, not loop recomputation; Loop is at best a partial intermediate. The affordance survives coarsening the store (N=128 to N=13): propagation attenuates but the memory/no-memory split persists, so fine granularity buys precision rather than the affordance itself. These results dissociate learning competence from editing affordance – repeated shared access suffices to grok, but edit propagation depends on whether the substrate exposes an addressable memory that the forward computation can write to and later reread, an affordance that loop recurrence provides only partially.
Read the full abstract

Artificial Intelligence · Cross-listed from cs.CV

Robust Zero-Shot Generalization for Open-Vocabulary Action Recognition via Task Arithmetic

Francesca Morandi, Omayma Moussadek, Federico Venturini et al. (+5 more)

Open Vocabulary Action Recognition (OVAR) enables the recognition of novel actions by leveraging vision-language representations, overcoming the limitations of traditional closed-set approaches. However, achieving robust performance in real-world scenarios typically requires domain-specific fine-tuning, which is often costly and raises privacy and regulatory concerns. In this work, we propose an alternative paradigm that bypasses target-domain training and recombines knowledge from existing datasets and models. Leveraging model merging and task arithmetic, we extract and combine task vectors from models fine-tuned on diverse public OVAR datasets. We show that, in out-of-distribution settings, the resulting merged model achieves superior zero-shot generalization to the pre-trained base model. Code is available at https://github.com/omaymaMoussadek/robust-ovar
Read the full abstract

All authors Francesca Morandi, Omayma Moussadek, Federico Venturini, Mauro Suardi, Alessandro Banzatti, Francesco Cannarile, Angelo Porrello, Simone Calderara

Artificial Intelligence · Cross-listed from cs.CV

XmoPipe: A Pipeline for Large-Scale In-the-Wild Human Motion Dataset Construction

Nathan Salazar, Emmanuel Dellandréa, Mathieu Lefort et al. (+1 more)

Large-scale human motion datasets are essential for training robust motion models for analysis, synthesis, and understanding. While marker-based motion capture provides precise data, it is costly and limited in scale and diversity. Recent advances in monocular motion capture and video-language understanding open the way to extract plausible motion from unconstrained online videos. We present a scalable pipeline for constructing in-the-wild human motion datasets. From a few keywords, the system retrieves videos, extracts 3D body and facial motion, and generates high-level textual descriptions. The pipeline is flexible, enabling targeted collection of various motions, multi-person interactions, or expressive behaviors. We demonstrate its quality by training motion reconstruction and motion generation models, showing performance comparable to models trained on traditional motion capture datasets and strong cross-dataset generalization.
Read the full abstract

All authors Nathan Salazar, Emmanuel Dellandréa, Mathieu Lefort, Alexandre Meyer

Artificial Intelligence · Cross-listed from physics.chem-ph

LLM-Guided Test-Time Discovery of Quantum-Chemical Approximation Algorithms

Masaya Hagai, Yuta Suzuki, Tomoya Murata et al. (+2 more)

Quantum chemistry simulations underpin modern materials discovery, yet their impact is limited by steep computational cost and dependence on fixed approximation schemes. Foundation models, such as machine-learned interatomic potentials, have accelerated parts of this workflow, but their reliance on large-scale pretraining restricts adaptability at the frontier of chemical space, where methodological innovation and sparse data are the norm. Agentic AI systems can automate existing simulation pipelines, yet they remain constrained by the predefined tools and algorithms they orchestrate. In response, we introduce LADeQ, an LLM-guided workflow that discovers, implements, and benchmarks candidate approximation algorithms at test-time within existing quantum chemistry codes. Rather than selecting from a predefined repertoire, LADeQ constructs candidate approximation schemes on demand, drawing on techniques from disciplines such as spatial statistics, circuit simulation, and kernel methods that have had little prior presence in electronic-structure theory. Because it builds on an out-of-the-box language model, LADeQ requires no task-specific pretraining or curated data, and the resulting implementations are transparent and inspectable, with explicitly traceable approximation errors that enable principled control of accuracy–efficiency trade-offs. We show that LADeQ accelerates coupled cluster singles and doubles (CCSD) and configuration interaction singles and doubles (CISD) calculations while keeping correlation-energy errors within user-specified tolerances, demonstrating autonomous, objective-driven discovery of approximation algorithms inside existing electronic-structure solvers.
Read the full abstract

All authors Masaya Hagai, Yuta Suzuki, Tomoya Murata, Shuhei Kurita, Masaki Adachi

Artificial Intelligence

Denoising Implicit Feedback for Cold-start Recommendation

Gaode Chen, Shicheng Wang, Shikun Li et al. (+8 more)

Implicit feedback is widely used in recommender systems due to its accessibility and generality, yet it usually presents noisy samples (e.g., clickbait, position bias). Meanwhile, recommenders inevitably face the item cold-start problem due to the continuous influx of new items. We identify that cold items are more prone to noisy samples due to the aforementioned factors, and researchers often overlook the significance of denoising implicit feedback for cold items. Previous denoising studies usually identify noisy samples based on heuristic patterns, such as higher loss values, and mitigate noise through sample selection or re-weighting. However, these methods have limited adaptability and are ineffective in cold-start scenarios. To achieve denoising implicit feedback for cold-start recommendation, we propose a model-agnostic denoising method called DIF. First, user preferences for content remain stable, which allows us to infer pseudo-labels indicating whether a user is interested in a cold item through content-similar warm items. Furthermore, to improve pseudo-label accuracy, we model the confidence of pseudo-labels based on the content similarity between the cold item and warm items, and then aggregate multiple pseudo-labels for each sample. Finally, we explicitly estimate the uncertainty of the noisy sample label by considering its relative entropy and the cold-start status of the item, which adaptively guides the role of pseudo-labels to correct the noisy labels at the sample level. DIF's superiority is supported by both theoretical justification and extensive experiments on real-world datasets. The method has been deployed on a billion-user scale short video application Kuaishou and has significantly improved various commercial metrics within cold-start scenarios.
Read the full abstract

All authors Gaode Chen, Shicheng Wang, Shikun Li, Rui Huang, Xinghua Zhang, Yunze Luo, Shipeng Li, Shiming Ge, Ruina Sun, Yinjie Jiang, Jun Zhang

Artificial Intelligence

BrainG3N: A Dual-Purpose Tokenizer for Controllable 3D Brain MRI Generation

Max Van Puyvelde, Ibrahim Gulluk, Wim Van Criekinge et al. (+1 more)

Three-dimensional (3D) brain MRI is central to clinical neurology and neuro-oncology, where generative models could augment under-represented cohorts, simulate disease trajectories, and support privacy-preserving data sharing. Latent diffusion has been the go-to solution for modeling imaging data, but it places two competing demands on the tokenizer: encoder embeddings must retain the clinical information that downstream tasks act on, and the decoder must reconstruct anatomically faithful volumes. Existing reconstruction-driven tokenizers achieve the second at the expense of the first. To address this, we introduce a fully volumetric masked-autoencoder (MAE) based tokenizer for 3D brain MRI latent diffusion, decoupling encoder and decoder: a frozen 3D MAE encoder produces clinically informative embeddings, while a dedicated CNN decoder reconstructs voxels from a linear projection of those embeddings. We pretrain the encoder on 35,309 volumes from 18 public cohorts spanning four modalities, ten disease categories, and 200+ acquisition sites, and demonstrate its dual utility in two settings. First, on a 23-task linear-probing benchmark, the encoder outperforms or matches SOTA models (i.e., BrainIAC, BrainSegFounder, and MedicalNet) on 21 of 23 tasks. Second, a conditional diffusion transformer (DiT) trained on these clinically informative embeddings supports both conditional generation across six variables and patient-specific longitudinal forecasting. Together these results establish a single 3D brain-MRI embedding space capable of both downstream clinical tasks and controllable generation.
Read the full abstract

All authors Max Van Puyvelde, Ibrahim Gulluk, Wim Van Criekinge, Olivier Gevaert

Artificial Intelligence · Cross-listed from cs.CL

Creating Multilingual Mental Health Dialogue Datasets: Limits of Persona-Based Localization via Nationality and Language

Yunkai Xu and Saeed Abdullah

AI and large language models (LLMs) have emerged as promising tools to address global mental health challenges. Despite the global nature of these challenges, there remains a critical shortage of high-quality datasets for training and evaluating such systems. To mitigate this gap, researchers increasingly generate synthetic clinical personas to simulate user data and test digital mental health support systems. However, most validated personas rely on English-centric contexts. This paper investigates whether similar persona-based methods can be used to generate multilingual mental health datasets. We modified nationality and language parameters in personas to generate clinical dialogues in Mandarin, Bengali, and Hindi. We then examined how different LLMs perform when evaluating the depression severity of these generated multilingual datasets against the baseline in English. Our findings indicate that just adding nationality and language parameters in personas might not be adequate, as it can introduce clinical inconsistency across languages. LLM judge models often exhibit inaccuracies in assessing depression severity in non-English texts, with performance varying across different models. This exposes the systemic limitations of applying English-centric personas to multilingual contexts. Ultimately, our work highlights the urgent need for culturally responsive data generation to ensure equitable mental health systems globally.
Read the full abstract