# Research Papers — 2026-01 · bundle page 107 of 189

> Full-text markdown of papers #5301–5350 of 9403 published in 2026-01, newest first. AI-readable mirrors of the full 9403-paper 2026-01 archive (50 papers per file).

Canonical page: https://aipod.dev/papers/archive/2026-01/page/107
[← Previous bundle](https://aipod.dev/papers/archive/2026-01/page/106.md)
[Next bundle →](https://aipod.dev/papers/archive/2026-01/page/108.md)

### Unleashing the Capabilities of Large Vision-Language Models for Intelligent Perception of Roadside Infrastructure

Authors: Luxuan Fu, Chong Liu, Bisheng Yang, Zhen Dong
Published: 2026-01-15
Categories: cs.CV
arXiv: https://arxiv.org/abs/2601.10551v1

Automated perception of urban roadside infrastructure is crucial for smart city management, yet general-purpose models often struggle to capture the necessary fine-grained attributes and domain rules. While Large Vision Language Models (VLMs) excel at open-world recognition, they often struggle to accurately interpret complex facility states in compliance with engineering standards, leading to unreliable performance in real-world applications. To address this, we propose a domain-adapted framework that transforms VLMs into specialized agents for intelligent infrastructure analysis. Our approach integrates a data-efficient fine-tuning strategy with a knowledge-grounded reasoning mechanism. Specifically, we leverage open-vocabulary fine-tuning on Grounding DINO to robustly localize diverse assets with minimal supervision, followed by LoRA-based adaptation on Qwen-VL for deep semantic attribute reasoning. To mitigate hallucinations and enforce professional compliance, we introduce a dual-modality Retrieval-Augmented Generation (RAG) module that dynamically retrieves authoritative industry standards and visual exemplars during inference. Evaluated on a comprehensive new dataset of urban roadside scenes, our framework achieves a detection performance of 58.9 mAP and an attribute recognition accuracy of 95.5%, demonstrating a robust solution for intelligent infrastructure monitoring.

---

### Defending Large Language Models Against Jailbreak Attacks via In-Decoding Safety-Awareness Probing

Authors: Yinzhi Zhao, Ming Wang, Shi Feng, Xiaocui Yang, Daling Wang, Yifei Zhang
Published: 2026-01-15
Categories: cs.AI, cs.CL
arXiv: https://arxiv.org/abs/2601.10543v2

Large language models (LLMs) have achieved impressive performance across natural language tasks and are increasingly deployed in real-world applications. Despite extensive safety alignment efforts, recent studies show that such alignment is often shallow and remains vulnerable to jailbreak attacks. Existing defense mechanisms, including decoding-based constraints and post-hoc content detectors, struggle against sophisticated jailbreaks, often intervening robust detection or excessively degrading model utility. In this work, we examine the decoding process of LLMs and make a key observation: even when successfully jailbroken, models internally exhibit latent safety-related signals during generation. However, these signals are overridden by the model's drive for fluent continuation, preventing timely self-correction or refusal. Building on this observation, we propose a simple yet effective approach that explicitly surfaces and leverages these latent safety signals for early detection of unsafe content during decoding. Experiments across diverse jailbreak attacks demonstrate that our approach significantly enhances safety, while maintaining low over-refusal rates on benign inputs and preserving response quality. Our results suggest that activating intrinsic safety-awareness during decoding offers a promising and complementary direction for defending against jailbreak attacks. Code is available at: https://github.com/zyz13590/SafeProbing.

---

### Mixtures of Transparent Local Models

Authors: Niffa Cheick Oumar Diaby, Thierry Duchesne, Mario Marchand
Published: 2026-01-15
Categories: cs.LG
arXiv: https://arxiv.org/abs/2601.10541v1

The predominance of machine learning models in many spheres of human activity has led to a growing demand for their transparency. The transparency of models makes it possible to discern some factors, such as security or non-discrimination. In this paper, we propose a mixture of transparent local models as an alternative solution for designing interpretable (or transparent) models. Our approach is designed for the situations where a simple and transparent function is suitable for modeling the label of instances in some localities/regions of the input space, but may change abruptly as we move from one locality to another. Consequently, the proposed algorithm is to learn both the transparent labeling function and the locality of the input space where the labeling function achieves a small risk in its assigned locality. By using a new multi-predictor (and multi-locality) loss function, we established rigorous PAC-Bayesian risk bounds for the case of binary linear classification problem and that of linear regression. In both cases, synthetic data sets were used to illustrate how the learning algorithms work. The results obtained from real data sets highlight the competitiveness of our approach compared to other existing methods as well as certain opaque models. Keywords: PAC-Bayes, risk bounds, local models, transparent models, mixtures of local transparent models.

---

### Enhancing the quality of gauge images captured in smoke and haze scenes through deep learning

Authors: Oscar H. Ramírez-Agudelo, Akshay N. Shewatkar, Edoardo Milana, Roland C. Aydin, Kai Franke
Published: 2026-01-15
Categories: cs.CV
arXiv: https://arxiv.org/abs/2601.10537v1

Images captured in hazy and smoky environments suffer from reduced visibility, posing a challenge when monitoring infrastructures and hindering emergency services during critical situations. The proposed work investigates the use of the deep learning models to enhance the automatic, machine-based readability of gauge in smoky environments, with accurate gauge data interpretation serving as a valuable tool for first responders. The study utilizes two deep learning architectures, FFA-Net and AECR-Net, to improve the visibility of gauge images, corrupted with light up to dense haze and smoke. Since benchmark datasets of analog gauge images are unavailable, a new synthetic dataset, containing over 14,000 images, was generated using the Unreal Engine. The models were trained with an 80\% train, 10\% validation, and 10\% test split for the haze and smoke dataset, respectively. For the synthetic haze dataset, the SSIM and PSNR metrics are about 0.98 and 43\,dB, respectively, comparing well to state-of-the art results. Additionally, more robust results are retrieved from the AECR-Net, when compared to the FFA-Net. Although the results from the synthetic smoke dataset are poorer, the trained models achieve interesting results. In general, imaging in the presence of smoke are more difficult to enhance given the inhomogeneity and high density. Secondly, FFA-Net and AECR-Net are implemented to dehaze and not to desmoke images. This work shows that use of deep learning architectures can improve the quality of analog gauge images captured in smoke and haze scenes immensely. Finally, the enhanced output images can be successfully post-processed for automatic autonomous reading of gauges

---

### CoGen: Creation of Reusable UI Components in Figma via Textual Commands

Authors: Ishani Kanapathipillai, Obhasha Priyankara
Published: 2026-01-15
Categories: cs.HC, cs.LG
arXiv: https://arxiv.org/abs/2601.10536v1

The evolution of User Interface design has emphasized the need for efficient, reusable, and editable components to ensure an efficient design process. This research introduces CoGen, a system that uses machine learning techniques to generate reusable UI components directly in Figma, one of the most popular UI design tools. Addressing gaps in current systems, CoGen focuses on creating atomic components such as buttons, labels, and input fields using structured JSON and natural language prompts. The project integrates Figma API data extraction, Seq2Seq models, and fine-tuned T5 transformers for component generation. The key results demonstrate the efficiency of the T5 model in prompt generation, with an accuracy of 98% and a BLEU score of 0.2668, which ensures the mapping of JSON to descriptive prompts. For JSON creation, CoGen achieves a success rate of up to 100% in generating simple JSON outputs for specified component types.

---

### SVII-3D: Advancing Roadside Infrastructure Inventory with Decimeter-level 3D Localization and Comprehension from Sparse Street Imagery

Authors: Chong Liu, Luxuan Fu, Yang Jia, Zhen Dong, Bisheng Yang
Published: 2026-01-15
Categories: cs.CV
arXiv: https://arxiv.org/abs/2601.10535v1

The automated creation of digital twins and precise asset inventories is a critical task in smart city construction and facility lifecycle management. However, utilizing cost-effective sparse imagery remains challenging due to limited robustness, inaccurate localization, and a lack of fine-grained state understanding. To address these limitations, SVII-3D, a unified framework for holistic asset digitization, is proposed. First, LoRA fine-tuned open-set detection is fused with a spatial-attention matching network to robustly associate observations across sparse views. Second, a geometry-guided refinement mechanism is introduced to resolve structural errors, achieving precise decimeter-level 3D localization. Third, transcending static geometric mapping, a Vision-Language Model agent leveraging multi-modal prompting is incorporated to automatically diagnose fine-grained operational states. Experiments demonstrate that SVII-3D significantly improves identification accuracy and minimizes localization errors. Consequently, this framework offers a scalable, cost-effective solution for high-fidelity infrastructure digitization, effectively bridging the gap between sparse perception and automated intelligent maintenance.

---

### PERM: Psychology-grounded Empathetic Reward Modeling for Large Language Models

Authors: Chengbing Wang, Wuqiang Zheng, Yang Zhang, Fengbin Zhu, Junyi Cheng, Yi Xie, Wenjie Wang, Fuli Feng
Published: 2026-01-15
Categories: cs.CL
arXiv: https://arxiv.org/abs/2601.10532v2

Large Language Models (LLMs) are increasingly deployed in human-centric applications, yet they often fail to provide substantive emotional support. While Reinforcement Learning (RL) has been utilized to enhance empathy of LLMs, existing reward models typically evaluate empathy from a single perspective, overlooking the inherently bidirectional interaction nature of empathy between the supporter and seeker as defined by Empathy Cycle theory. To address this limitation, we propose Psychology-grounded Empathetic Reward Modeling (PERM). PERM operationalizes empathy evaluation through a bidirectional decomposition: 1) Supporter perspective, assessing internal resonation and communicative expression; 2) Seeker perspective, evaluating emotional reception. Additionally, it incorporates a bystander perspective to monitor overall interaction quality. Extensive experiments on a widely-used emotional intelligence benchmark and an industrial daily conversation dataset demonstrate that PERM outperforms state-of-the-art baselines by over 10\%. Furthermore, a blinded user study reveals a 70\% preference for our approach, highlighting its efficacy in generating more empathetic responses. Our code, dataset, and models are available at https://github.com/ZhengWwwq/PERM.

---

### Coarsening Causal DAG Models

Authors: Francisco Madaleno, Pratik Misra, Alex Markham
Published: 2026-01-15
Categories: stat.ML, cs.LG, math.CO
arXiv: https://arxiv.org/abs/2601.10531v2

Directed acyclic graphical (DAG) models are a powerful tool for representing causal relationships among jointly distributed random variables, especially concerning data from across different experimental settings. However, it is not always practical or desirable to estimate a causal model at the granularity of given features in a particular dataset. There is a growing body of research on causal abstraction to address such problems. We contribute to this line of research by (i) providing novel graphical identifiability results for practically-relevant interventional settings, (ii) proposing an efficient, provably consistent algorithm for directly learning abstract causal graphs from interventional data with unknown intervention targets, and (iii) uncovering theoretical insights about the lattice structure of the underlying search space, with connections to the field of causal discovery more generally. As proof of concept, we apply our algorithm on synthetic and real datasets with known ground truths, including measurements from a controlled physical system with interacting light intensity and polarization.

---

### A Safety Report on GPT-5.2, Gemini 3 Pro, Qwen3-VL, Grok 4.1 Fast, Nano Banana Pro, and Seedream 4.5

Authors: Xingjun Ma, Yixu Wang, Hengyuan Xu, Yutao Wu, Yifan Ding, Yunhan Zhao, Zilong Wang, Jiabin Hua, Ming Wen, Jianan Liu, Ranjie Duan, Yifeng Gao et al.
Published: 2026-01-15
Categories: cs.AI, cs.CL, cs.CV, cs.LG
arXiv: https://arxiv.org/abs/2601.10527v2

The rapid evolution of Large Language Models (LLMs) and Multimodal Large Language Models (MLLMs) has driven major gains in reasoning, perception, and generation across language and vision, yet whether these advances translate into comparable improvements in safety remains unclear, partly due to fragmented evaluations that focus on isolated modalities or threat models. In this report, we present an integrated safety evaluation of six frontier models--GPT-5.2, Gemini 3 Pro, Qwen3-VL, Grok 4.1 Fast, Nano Banana Pro, and Seedream 4.5--assessing each across language, vision-language, and image generation using a unified protocol that combines benchmark, adversarial, multilingual, and compliance evaluations. By aggregating results into safety leaderboards and model profiles, we reveal a highly uneven safety landscape: while GPT-5.2 demonstrates consistently strong and balanced performance, other models exhibit clear trade-offs across benchmark safety, adversarial robustness, multilingual generalization, and regulatory compliance. Despite strong results under standard benchmarks, all models remain highly vulnerable under adversarial testing, with worst-case safety rates dropping below 6%. Text-to-image models show slightly stronger alignment in regulated visual risk categories, yet remain fragile when faced with adversarial or semantically ambiguous prompts. Overall, these findings highlight that safety in frontier models is inherently multidimensional--shaped by modality, language, and evaluation design--underscoring the need for standardized, holistic safety assessments to better reflect real-world risk and guide responsible deployment.

---

### Diagnosing Generalization Failures in Fine-Tuned LLMs: A Cross-Architectural Study on Phishing Detection

Authors: Frank Bobe, Gregory D. Vetaw, Chase Pavlick, Darshan Bryner, Matthew Cook, Jose Salas-Vernis
Published: 2026-01-15
Categories: cs.AI
arXiv: https://arxiv.org/abs/2601.10524v1

The practice of fine-tuning Large Language Models (LLMs) has achieved state-of-the-art performance on specialized tasks, yet diagnosing why these models become brittle and fail to generalize remains a critical open problem. To address this, we introduce and apply a multi-layered diagnostic framework to a cross-architectural study. We fine-tune Llama 3.1 8B, Gemma 2 9B, and Mistral models on a high-stakes phishing detection task and use SHAP analysis and mechanistic interpretability to uncover the root causes of their generalization failures. Our investigation reveals three critical findings: (1) Generalization is driven by a powerful synergy between architecture and data diversity. The Gemma 2 9B model achieves state-of-the-art performance (>91\% F1), but only when trained on a stylistically diverse ``generalist'' dataset. (2) Generalization is highly architecture-dependent. We diagnose a specific failure mode in Llama 3.1 8B, which performs well on a narrow domain but cannot integrate diverse data, leading to a significant performance drop. (3) Some architectures are inherently more generalizable. The Mistral model proves to be a consistent and resilient performer across multiple training paradigms. By pinpointing the flawed heuristics responsible for these failures, our work provides a concrete methodology for diagnosing and understanding generalization failures, underscoring that reliable AI requires deep validation of the interplay between architecture, data, and training strategy.

---

### BikeActions: An Open Platform and Benchmark for Cyclist-Centric VRU Action Recognition

Authors: Max A. Buettner, Kanak Mazumder, Luca Koecher, Mario Finkbeiner, Sebastian Niebler, Fabian B. Flohr
Published: 2026-01-15
Categories: cs.CV
arXiv: https://arxiv.org/abs/2601.10521v2

Anticipating the intentions of Vulnerable Road Users (VRUs) is a critical challenge for safe autonomous driving (AD) and mobile robotics. While current research predominantly focuses on pedestrian crossing behaviors from a vehicle's perspective, interactions within dense shared spaces remain underexplored. To bridge this gap, we introduce FUSE-Bike, the first fully open perception platform of its kind. Equipped with two LiDARs, a camera, and GNSS, it facilitates high-fidelity, close-range data capture directly from a cyclist's viewpoint. Leveraging this platform, we present BikeActions, a novel multi-modal dataset comprising 852 annotated samples across 5 distinct action classes, specifically tailored to improve VRU behavior modeling. We establish a rigorous benchmark by evaluating state-of-the-art graph convolution and transformer-based models on our publicly released data splits, establishing the first performance baselines for this challenging task. We release the full dataset together with data curation tools, the open hardware design, and the benchmark code to foster future research in VRU action understanding under https://iv.ee.hm.edu/bikeactions/.

---

### Breaking Up with Normatively Monolithic Agency with GRACE: A Reason-Based Neuro-Symbolic Architecture for Safe and Ethical AI Alignment

Authors: Felix Jahn, Yannic Muskalla, Lisa Dargasz, Patrick Schramowski, Kevin Baum
Published: 2026-01-15
Categories: cs.AI, cs.CY
arXiv: https://arxiv.org/abs/2601.10520v2

As AI agents become increasingly autonomous, widely deployed in consequential contexts, and efficacious in bringing about real-world impacts, ensuring that their decisions are not only instrumentally effective but also normatively aligned has become critical. We introduce a neuro-symbolic reason-based containment architecture, Governor for Reason-Aligned ContainmEnt (GRACE), that decouples normative reasoning from instrumental decision-making and can contain AI agents of virtually any design. GRACE restructures decision-making into three modules: a Moral Module (MM) that determines permissible macro actions via deontic logic-based reasoning; a Decision-Making Module (DMM) that encapsulates the target agent while selecting instrumentally optimal primitive actions in accordance with derived macro actions; and a Guard that monitors and enforces moral compliance. The MM uses a reason-based formalism providing a semantic foundation for deontic logic, enabling interpretability, contestability, and justifiability. Its symbolic representation enriches the DMM's informational context and supports formal verification and statistical guarantees of alignment enforced by the Guard. We demonstrate GRACE on an example of a LLM therapy assistant, showing how it enables stakeholders to understand, contest, and refine agent behavior.

---

### Transformer-Based Cognitive Radio: Adaptive Modulation Strategies Using Transformer Models

Authors: Andrea Melis, Andrea Piroddi, Roberto Girau
Published: 2026-01-15
Categories: cs.LG
arXiv: https://arxiv.org/abs/2601.10519v1

Cognitive Radio (CR) systems, which dynamically adapt to changing spectrum environments, could benefit significantly from advancements in machine learning technologies. These systems can be enhanced in terms of spectral efficiency, robustness, and security through innovative approaches such as the use of Transformer models. This work investigates the application of Transformer models, specifically the GPT-2 architecture, to generate novel modulation schemes for wireless communications. By training a GPT-2 model on a dataset of existing modulation formulas, new modulation schemes has been created. These generated schemes are then compared to traditional methods using key performance metrics such as Signal-to-Noise Ratio (SNR) and Power Spectrum Density (PSD). The results show that Transformer-generated modulation schemes can achieve performance comparable to, and in some cases outperforming, traditional methods. This demonstrates that advanced CR systems could greatly benefit from the implementation of Transformer models, leading to more efficient, robust, and secure communication systems.

---

### AEQ-Bench: Measuring Empathy of Omni-Modal Large Models

Authors: Xuan Luo, Lewei Yao, Libo Zhao, Lanqing Hong, Kai Chen, Dehua Tao, Daxin Tan, Ruifeng Xu, Jing Li
Published: 2026-01-15
Categories: cs.CL, cs.HC
arXiv: https://arxiv.org/abs/2601.10513v1

While the automatic evaluation of omni-modal large models (OLMs) is essential, assessing empathy remains a significant challenge due to its inherent affectivity. To investigate this challenge, we introduce AEQ-Bench (Audio Empathy Quotient Benchmark), a novel benchmark to systematically assess two core empathetic capabilities of OLMs: (i) generating empathetic responses by comprehending affective cues from multi-modal inputs (audio + text), and (ii) judging the empathy of audio responses without relying on text transcription. Compared to existing benchmarks, AEQ-Bench incorporates two novel settings that vary in context specificity and speech tone. Comprehensive assessment across linguistic and paralinguistic metrics reveals that (1) OLMs trained with audio output capabilities generally outperformed models with text-only outputs, and (2) while OLMs align with human judgments for coarse-grained quality assessment, they remain unreliable for evaluating fine-grained paralinguistic expressiveness.

---

### SatMap: Revisiting Satellite Maps as Prior for Online HD Map Construction

Authors: Kanak Mazumder, Fabian B. Flohr
Published: 2026-01-15
Categories: cs.CV, cs.AI, cs.LG
arXiv: https://arxiv.org/abs/2601.10512v2

Online high-definition (HD) map construction is an essential part of a safe and robust end-to-end autonomous driving (AD) pipeline. Onboard camera-based approaches suffer from limited depth perception and degraded accuracy due to occlusion. In this work, we propose SatMap, an online vectorized HD map estimation method that integrates satellite maps with multi-view camera observations and directly predicts a vectorized HD map for downstream prediction and planning modules. Our method leverages lane-level semantics and texture from satellite imagery captured from a Bird's Eye View (BEV) perspective as a global prior, effectively mitigating depth ambiguity and occlusion. In our experiments on the nuScenes dataset, SatMap achieves 34.8% mAP performance improvement over the camera-only baseline and 8.5% mAP improvement over the camera-LiDAR fusion baseline. Moreover, we evaluate our model in long-range and adverse weather conditions to demonstrate the advantages of using a satellite prior map. Source code will be available at https://iv.ee.hm.edu/satmap/.

---

### Scalable Algorithms for Approximate DNF Model Counting

Authors: Paul Burkhardt, David G. Harris, Kevin T Schmitt
Published: 2026-01-15
Categories: cs.DS, cs.AI
arXiv: https://arxiv.org/abs/2601.10511v2

Model counting of Disjunctive Normal Form (DNF) formulas is a critical problem in applications such as probabilistic inference and network reliability. For example, it is often used for query evaluation in probabilistic databases. Due to the computational intractability of exact DNF counting, there has been a line of research into a variety of approximation algorithms. These include Monte Carlo approaches such as the classical algorithms of Karp, Luby, and Madras (1989), as well as methods based on hashing (Soos et al. 2023), and heuristic approximations based on Neural Nets (Abboud, Ceylan, and Lukasiewicz 2020). We develop a new Monte Carlo approach with an adaptive stopping rule and short-circuit formula evaluation. We prove it achieves Probably Approximately Correct (PAC) learning bounds and is asymptotically more efficient than the previous methods. We also show experimentally that it out-performs prior algorithms by orders of magnitude, and can scale to much larger problems with millions of variables.

---

### DR-Arena: an Automated Evaluation Framework for Deep Research Agents

Authors: Yiwen Gao, Ruochen Zhao, Yang Deng, Wenxuan Zhang
Published: 2026-01-15
Categories: cs.CL
arXiv: https://arxiv.org/abs/2601.10504v3

As Large Language Models (LLMs) increasingly operate as Deep Research (DR) Agents capable of autonomous investigation and information synthesis, reliable evaluation of their task performance has become a critical bottleneck. Current benchmarks predominantly rely on static datasets, which suffer from several limitations: limited task generality, temporal misalignment, and data contamination. To address these, we introduce DR-Arena, a fully automated evaluation framework that pushes DR agents to their capability limits through dynamic investigation. DR-Arena constructs real-time Information Trees from fresh web trends to ensure the evaluation rubric is synchronized with the live world state, and employs an automated Examiner to generate structured tasks testing two orthogonal capabilities: Deep reasoning and Wide coverage. DR-Arena further adopts Adaptive Evolvement Loop, a state-machine controller that dynamically escalates task complexity based on real-time performance, demanding deeper deduction or wider aggregation until a decisive capability boundary emerges. Experiments with six advanced DR agents demonstrate that DR-Arena achieves a Spearman correlation of 0.94 with the LMSYS Search Arena leaderboard. This represents the state-of-the-art alignment with human preferences without any manual efforts, validating DR-Arena as a reliable alternative for costly human adjudication.

---

### PROMA: Projected Microbatch Accumulation for Reference-Free Proximal Policy Updates

Authors: Nilin Abrahamsen
Published: 2026-01-15
Categories: cs.LG, cs.AI
arXiv: https://arxiv.org/abs/2601.10498v4

This note introduces Projected Microbatch Accumulation (PROMA), a reference-free proximal policy method that controls KL divergence by projecting away high-variance components of the policy gradient. Two variants are presented. In the accumulation-based variant, the running gradient is projected orthogonal to the sequence-wise log-probability gradients of each microbatch. In the intra-microbatch variant, a factored projection using dominant subspaces of activations and gradient outputs is applied independently within each microbatch, making it compatible with standard data-parallel training. Empirically, the accumulation variant achieves tighter per-step KL control than GRPO with PPO clipping, while the intra-microbatch variant achieves the best validation performance.

---

### MERGETUNE: Continued Fine-Tuning of Vision-Language Models

Authors: Wenqing Wang, Da Li, Xiatian Zhu, Josef Kittler
Published: 2026-01-15
Categories: cs.CV
arXiv: https://arxiv.org/abs/2601.10497v3

Fine-tuning vision-language models (VLMs) such as CLIP often leads to catastrophic forgetting of pretrained knowledge. Prior work primarily aims to mitigate forgetting during adaptation; however, forgetting often remains inevitable during this process. We introduce a novel paradigm, continued fine-tuning (CFT), which seeks to recover pretrained knowledge after a zero-shot model has already been adapted. We propose a simple, model-agnostic CFT strategy (named MERGETUNE) guided by linear mode connectivity (LMC), which can be applied post hoc to existing fine-tuned models without requiring architectural changes. Given a fine-tuned model, we continue fine-tuning its trainable parameters (e.g., soft prompts or linear heads) to search for a continued model which has two low-loss paths to the zero-shot (e.g., CLIP) and the fine-tuned (e.g., CoOp) solutions. By exploiting the geometry of the loss landscape, the continued model implicitly merges the two solutions, restoring pretrained knowledge lost in the fine-tuned counterpart. A challenge is that the vanilla LMC constraint requires data replay from the pretraining task. We approximate this constraint for the zero-shot model via a second-order surrogate, eliminating the need for large-scale data replay. Experiments show that MERGETUNE improves the harmonic mean of CoOp by +5.6% on base-novel generalisation without adding parameters. On robust fine-tuning evaluations, the LMC-merged model from MERGETUNE surpasses ensemble baselines with lower inference cost, achieving further gains and state-of-the-art results when ensembled with the zero-shot model. Our code is available at https://github.com/Surrey-UP-Lab/MERGETUNE.

---

### Model See, Model Do? Exposure-Aware Evaluation of Bug-vs-Fix Preference in Code LLMs

Authors: Ali Al-Kaswan, Claudio Spiess, Prem Devanbu, Arie van Deursen, Maliheh Izadi
Published: 2026-01-15
Categories: cs.SE, cs.AI
arXiv: https://arxiv.org/abs/2601.10496v1

Large language models are increasingly used for code generation and debugging, but their outputs can still contain bugs, that originate from training data. Distinguishing whether an LLM prefers correct code, or a familiar incorrect version might be influenced by what it's been exposed to during training. We introduce an exposure-aware evaluation framework that quantifies how prior exposure to buggy versus fixed code influences a model's preference. Using the ManySStuBs4J benchmark, we apply Data Portraits for membership testing on the Stack-V2 corpus to estimate whether each buggy and fixed variant was seen during training. We then stratify examples by exposure and compare model preference using code completion as well as multiple likelihood-based scoring metrics We find that most examples (67%) have neither variant in the training data, and when only one is present, fixes are more frequently present than bugs. In model generations, models reproduce buggy lines far more often than fixes, with bug-exposed examples amplifying this tendency and fix-exposed examples showing only marginal improvement. In likelihood scoring, minimum and maximum token-probability metrics consistently prefer the fixed code across all conditions, indicating a stable bias toward correct fixes. In contrast, metrics like the Gini coefficient reverse preference when only the buggy variant was seen. Our results indicate that exposure can skew bug-fix evaluations and highlight the risk that LLMs may propagate memorised errors in practice.

---

### CROCS: A Two-Stage Clustering Framework for Behaviour-Centric Consumer Segmentation with Smart Meter Data

Authors: Luke W. Yerbury, Ricardo J. G. B. Campello, G. C. Livingston, Mark Goldsworthy, Lachlan O'Neil
Published: 2026-01-15
Categories: stat.ML, cs.LG
arXiv: https://arxiv.org/abs/2601.10494v3

With grid operators confronting rising uncertainty from renewable integration and a broader push toward electrification, Demand-Side Management (DSM) -- particularly Demand Response (DR) -- has attracted significant attention as a cost-effective mechanism for balancing modern electricity systems. Unprecedented volumes of consumption data from a continuing global deployment of smart meters enable consumer segmentation based on real usage behaviours, promising to inform the design of more effective DSM and DR programs. However, existing clustering-based segmentation methods insufficiently reflect the behavioural diversity of consumers, often relying on rigid temporal alignment, and faltering in the presence of anomalies, missing data, or large-scale deployments. To address these challenges, we propose a novel two-stage clustering framework -- Clustered Representations Optimising Consumer Segmentation (CROCS). In the first stage, each consumer's daily load profiles are clustered independently to form a Representative Load Set (RLS), providing a compact summary of their typical diurnal consumption behaviours. In the second stage, consumers are clustered using the Weighted Sum of Minimum Distances (WSMD), a novel set-to-set measure that compares RLSs by accounting for both the prevalence and similarity of those behaviours. Finally, community detection on the WSMD-induced graph reveals higher-order prototypes that embody the shared diurnal behaviours defining consumer groups, enhancing the interpretability of the resulting clusters. Extensive experiments on both synthetic and real Australian smart meter datasets demonstrate that CROCS captures intra-consumer variability, uncovers both synchronous and asynchronous behavioural similarities, and remains robust to anomalies and missing data, while scaling efficiently through natural parallelisation. These results...

---

### Communication-Efficient Federated Learning by Exploiting Spatio-Temporal Correlations of Gradients

Authors: Shenlong Zheng, Zhen Zhang, Yuhui Deng, Geyong Min, Lin Cui
Published: 2026-01-15
Categories: cs.LG
arXiv: https://arxiv.org/abs/2601.10491v1

Communication overhead is a critical challenge in federated learning, particularly in bandwidth-constrained networks. Although many methods have been proposed to reduce communication overhead, most focus solely on compressing individual gradients, overlooking the temporal correlations among them. Prior studies have shown that gradients exhibit spatial correlations, typically reflected in low-rank structures. Through empirical analysis, we further observe a strong temporal correlation between client gradients across adjacent rounds. Based on these observations, we propose GradESTC, a compression technique that exploits both spatial and temporal gradient correlations. GradESTC exploits spatial correlations to decompose each full gradient into a compact set of basis vectors and corresponding combination coefficients. By exploiting temporal correlations, only a small portion of the basis vectors need to be dynamically updated in each round. GradESTC significantly reduces communication overhead by transmitting lightweight combination coefficients and a limited number of updated basis vectors instead of the full gradients. Extensive experiments show that, upon reaching a target accuracy level near convergence, GradESTC reduces uplink communication by an average of 39.79% compared to the strongest baseline, while maintaining comparable convergence speed and final accuracy to uncompressed FedAvg. By effectively leveraging spatio-temporal gradient structures, GradESTC offers a practical and scalable solution for communication-efficient federated learning.

---

### Panning for Gold: Expanding Domain-Specific Knowledge Graphs with General Knowledge

Authors: Runhao Zhao, Weixin Zeng, Wentao Zhang, Chong Chen, Zhengpin Li, Xiang Zhao, Lei Chen
Published: 2026-01-15
Categories: cs.AI
arXiv: https://arxiv.org/abs/2601.10485v5

Domain-specific knowledge graphs (DKGs) are critical yet often suffer from limited coverage compared to General Knowledge Graphs (GKGs). Existing tasks to enrich DKGs rely primarily on extracting knowledge from external unstructured data or completing KGs through internal reasoning, but the scope and quality of such integration remain limited. This highlights a critical gap: little systematic exploration has been conducted on how comprehensive, high-quality GKGs can be effectively leveraged to supplement DKGs. To address this gap, we propose a new and practical task: domain-specific knowledge graph fusion (DKGF), which aims to mine and integrate relevant facts from general knowledge graphs into domain-specific knowledge graphs to enhance their completeness and utility. Unlike previous research, this new task faces two key challenges: (1) high ambiguity of domain relevance, i.e., difficulty in determining whether knowledge from a GKG is truly relevant to the target domain , and (2) cross-domain knowledge granularity misalignment, i.e., GKG facts are typically abstract and coarse-grained, whereas DKGs frequently require more contextualized, fine-grained representations aligned with particular domain scenarios. To address these, we present ExeFuse, a neuro-symbolic framework based on a novel Fact-as-Program paradigm. ExeFuse treats fusion as an executable process, utilizing neuro-symbolic execution to infer logical relevance beyond surface similarity and employing target space grounding to calibrate granularity. We construct new datasets to establish the first standardized evaluation suite for this task. Extensive experiments demonstrate that ExeFuse effectively overcomes domain barriers to achieve superior fusion performance.

---

### High-Dimensional Analysis of Gradient Flow for Extensive-Width Quadratic Neural Networks

Authors: Simon Martin, Giulio Biroli, Francis Bach
Published: 2026-01-15
Categories: math.OC, cond-mat.dis-nn, stat.ML
arXiv: https://arxiv.org/abs/2601.10483v2

We study the high-dimensional training dynamics of a shallow neural network with quadratic activation in a teacher-student setup. We focus on the extensive-width regime, where the teacher and student network widths scale proportionally with the input dimension, and the sample size grows quadratically. This scaling aims to describe overparameterized neural networks in which feature learning still plays a central role. In the high-dimensional limit, we derive a dynamical characterization of the gradient flow, in the spirit of dynamical mean-field theory (DMFT). Under l2-regularization, we analyze these equations at long times and characterize the performance and spectral properties of the resulting estimator. This result provides a quantitative understanding of the effect of overparameterization on learning and generalization, and reveals a double descent phenomenon in the presence of label noise, where generalization improves beyond interpolation. In the small regularization limit, we obtain an exact expression for the perfect recovery threshold as a function of the network widths, providing a precise characterization of how overparameterization influences recovery.

---

### H-EFT-VA: An Effective-Field-Theory Variational Ansatz with Provable Barren Plateau Avoidance

Authors: Eyad I. B Hamid
Published: 2026-01-15
Categories: quant-ph, cs.LG, math-ph
arXiv: https://arxiv.org/abs/2601.10479v2

Variational Quantum Algorithms (VQAs) are critically threatened by the Barren Plateau (BP) phenomenon. In this work, we introduce the H-EFT Variational Ansatz (H-EFT-VA), an architecture inspired by Effective Field Theory (EFT). By enforcing a hierarchical "UV-cutoff" on initialization, we theoretically restrict the circuit's state exploration, preventing the formation of approximate unitary 2-designs. We provide a rigorous proof that this localization guarantees an inverse-polynomial lower bound on the gradient variance: $Var(\partialθ) \in Ω(1/poly(N))$. Crucially, unlike approaches that avoid BPs by limiting entanglement, we demonstrate that H-EFT-VA maintains volume-law entanglement and near-Haar purity, ensuring sufficient expressibility for complex quantum states. Extensive benchmarking across 16 experiments on the Transverse Field Ising Model confirms a 109x improvement in energy convergence and a 10.7x increase in ground-state fidelity over standard Hardware-Efficient Ansätze (HEA), with statistical significance of $p < 10^{-88}$. The static framework is most effective for Hamiltonians with moderate reference-state overlap; extension to systems with larger reference-state gaps is addressed through dynamic UV-cutoff relaxation strategies explored in concurrent work.

---

### Urban Socio-Semantic Segmentation with Vision-Language Reasoning

Authors: Yu Wang, Yi Wang, Rui Dai, Yujie Wang, Kaikui Liu, Xiangxiang Chu, Yansheng Li
Published: 2026-01-15
Categories: cs.CV, cs.AI, cs.CY
arXiv: https://arxiv.org/abs/2601.10477v2

As hubs of human activity, urban surfaces consist of a wealth of semantic entities. Segmenting these various entities from satellite imagery is crucial for a range of downstream applications. Current advanced segmentation models can reliably segment entities defined by physical attributes (e.g., buildings, water bodies) but still struggle with socially defined categories (e.g., schools, parks). In this work, we achieve socio-semantic segmentation by vision-language model reasoning. To facilitate this, we introduce the Urban Socio-Semantic Segmentation dataset named SocioSeg, a new resource comprising satellite imagery, digital maps, and pixel-level labels of social semantic entities organized in a hierarchical structure. Additionally, we propose a novel vision-language reasoning framework called SocioReasoner that simulates the human process of identifying and annotating social semantic entities via cross-modal recognition and multi-stage reasoning. We employ reinforcement learning to optimize this non-differentiable process and elicit the reasoning capabilities of the vision-language model. Experiments demonstrate our approach's gains over state-of-the-art models and strong zero-shot generalization. The dataset and code are open-sourced under the Apache License 2.0 at https://github.com/AMAP-ML/SocioReasoner.

---

### DeFlow: Decoupling Manifold Modeling and Value Maximization for Offline Policy Extraction

Authors: Zhancun Mu
Published: 2026-01-15
Categories: cs.LG
arXiv: https://arxiv.org/abs/2601.10471v2

We present DeFlow, a decoupled offline RL framework that leverages flow matching to faithfully capture complex behavior manifolds. Optimizing generative policies is computationally prohibitive, typically necessitating backpropagation through ODE solvers. We address this by learning a lightweight refinement module within an explicit, data-derived trust region of the flow manifold, rather than sacrificing the iterative generation capability via single-step distillation. This way, we bypass solver differentiation and eliminate the need for balancing loss terms, ensuring stable improvement while fully preserving the flow's iterative expressivity. Empirically, DeFlow achieves superior performance on the challenging OGBench benchmark and demonstrates efficient offline-to-online adaptation.

---

### ChartComplete: A Taxonomy-based Inclusive Chart Dataset

Authors: Ahmad Mustapha, Charbel Toumieh, Mariette Awad
Published: 2026-01-15
Categories: cs.AI, cs.CV
arXiv: https://arxiv.org/abs/2601.10462v3

With advancements in deep learning (DL) and computer vision techniques, the field of chart understanding is evolving rapidly. In particular, multimodal large language models (MLLMs) are proving to be efficient and accurate in understanding charts. To accurately measure the performance of MLLMs, the research community has developed multiple datasets to serve as benchmarks. By examining these datasets, we found that they are all limited to a small set of chart types. To bridge this gap, we propose the ChartComplete dataset. The dataset is based on a chart taxonomy borrowed from the visualization community, and it covers thirty different chart types. The dataset is a collection of classified chart images and does not include a learning signal. We present the ChartComplete dataset as is to the community to build upon it.

---

### Contextual StereoSet: Stress-Testing Bias Alignment Robustness in Large Language Models

Authors: Abhinaba Basu, Pavan Chakraborty
Published: 2026-01-15
Categories: cs.CL, cs.AI, cs.CY, cs.LG
arXiv: https://arxiv.org/abs/2601.10460v1

A model that avoids stereotypes in a lab benchmark may not avoid them in deployment. We show that measured bias shifts dramatically when prompts mention different places, times, or audiences -- no adversarial prompting required. We introduce Contextual StereoSet, a benchmark that holds stereotype content fixed while systematically varying contextual framing. Testing 13 models across two protocols, we find striking patterns: anchoring to 1990 (vs. 2030) raises stereotype selection in all models tested on this contrast (p<0.05); gossip framing raises it in 5 of 6 full-grid models; out-group observer framing shifts it by up to 13 percentage points. These effects replicate in hiring, lending, and help-seeking vignettes. We propose Context Sensitivity Fingerprints (CSF): a compact profile of per-dimension dispersion and paired contrasts with bootstrap CIs and FDR correction. Two evaluation tracks support different use cases -- a 360-context diagnostic grid for deep analysis and a budgeted protocol covering 4,229 items for production screening. The implication is methodological: bias scores from fixed-condition tests may not generalize.This is not a claim about ground-truth bias rates; it is a stress test of evaluation robustness. CSF forces evaluators to ask, "Under what conditions does bias appear?" rather than "Is this model biased?" We release our benchmark, code, and results.

---

### LangLasso: Interactive Cluster Descriptions through LLM Explanation

Authors: Raphael Buchmüller, Dennis Collaris, Linhao Meng, Angelos Chatzimparmpas
Published: 2026-01-15
Categories: cs.HC, cs.LG, stat.CO
arXiv: https://arxiv.org/abs/2601.10458v1

Dimensionality reduction is a powerful technique for revealing structure and potential clusters in data. However, as the axes are complex, non-linear combinations of features, they often lack semantic interpretability. Existing visual analytics (VA) methods support cluster interpretation through feature comparison and interactive exploration, but they require technical expertise and intense human effort. We present \textit{LangLasso}, a novel method that complements VA approaches through interactive, natural language descriptions of clusters using large language models (LLMs). It produces human-readable descriptions that make cluster interpretation accessible to non-experts and allow integration of external contextual knowledge beyond the dataset. We systematically evaluate the reliability of these explanations and demonstrate that \langlasso provides an effective first step for engaging broader audiences in cluster interpretation. The tool is available at https://langlasso.vercel.app

---

### NSR-Boost: A Neuro-Symbolic Residual Boosting Framework for Industrial Legacy Models

Authors: Ziming Dai, Dabiao Ma, Jinle Tong, Mengyuan Han, Jian Yang, Hongtao Liu, Haojun Fei, Qing Yang
Published: 2026-01-15
Categories: cs.AI
arXiv: https://arxiv.org/abs/2601.10457v3

Although the Gradient Boosted Decision Trees (GBDTs) dominate industrial tabular applications, upgrading legacy models in high-concurrency production environments still faces prohibitive retraining costs and systemic risks. To address this problem, we present NSR-Boost, a neuro-symbolic residual boosting framework designed specifically for industrial scenarios. Its core advantage lies in being ``non-intrusive''. It treats the legacy model as a frozen model and performs targeted repairs on "hard regions" where predictions fail. The framework comprises three key stages: First, finding hard regions through residuals, then generating interpretable experts by generating symbolic code structures using Large Language Model (LLM) and fine-tuning parameters using Bayesian optimization, and finally dynamically integrating experts with legacy model output through a lightweight aggregator. Experimental results demonstrate that the framework significantly outperforms state-of-the-art (SOTA) baselines across six public datasets and one private dataset. More importantly, we report the successful deployment of NSR-Boost within the core financial risk control system of Qfin Holdings, where empirical results on real-world online traffic exhibit superior performance improvements and a significant reduction in the bad rate. In conclusion, it effectively captures long-tail risks missed by traditional models and offers a safe, low-cost evolutionary paradigm for industry.

---

### SurgGoal: Rethinking Surgical Planning Evaluation via Goal-Satisfiability

Authors: Ruochen Li, Kun Yuan, Yufei Xia, Yue Zhou, Qingyu Lu, Weihang Li, Youxiang Zhu, Nassir Navab
Published: 2026-01-15
Categories: cs.CL, cs.RO
arXiv: https://arxiv.org/abs/2601.10455v1

Surgical planning integrates visual perception, long-horizon reasoning, and procedural knowledge, yet it remains unclear whether current evaluation protocols reliably assess vision-language models (VLMs) in safety-critical settings. Motivated by a goal-oriented view of surgical planning, we define planning correctness via phase-goal satisfiability, where plan validity is determined by expert-defined surgical rules. Based on this definition, we introduce a multicentric meta-evaluation benchmark with valid procedural variations and invalid plans containing order and content errors. Using this benchmark, we show that sequence similarity metrics systematically misjudge planning quality, penalizing valid plans while failing to identify invalid ones. We therefore adopt a rule-based goal-satisfiability metric as a high-precision meta-evaluation reference to assess Video-LLMs under progressively constrained settings, revealing failures due to perception errors and under-constrained reasoning. Structural knowledge consistently improves performance, whereas semantic guidance alone is unreliable and benefits larger models only when combined with structural constraints.

---

### Stable Differentiable Modal Synthesis for Learning Nonlinear Dynamics

Authors: Victor Zheleznov, Stefan Bilbao, Alec Wright, Simon King
Published: 2026-01-15
Categories: cs.SD, cs.LG, eess.AS, physics.comp-ph
arXiv: https://arxiv.org/abs/2601.10453v3

Modal methods are a long-standing approach to physical modelling synthesis. Extensions to nonlinear problems are possible, leading to coupled nonlinear systems of ordinary differential equations. Recent work in scalar auxiliary variable techniques has enabled construction of explicit and stable numerical solvers for such systems. On the other hand, neural ordinary differential equations have been successful in modelling nonlinear systems from data. In this work, we examine how scalar auxiliary variable techniques can be combined with neural ordinary differential equations to yield a stable differentiable model capable of learning nonlinear dynamics. The proposed approach leverages the analytical solution for linear vibration of the system's modes so that physical parameters of a system remain easily accessible after the training without the need for a parameter encoder in the model architecture. Compared to our previous work that used multilayer perceptrons to parametrise nonlinear dynamics, we employ gradient networks that allow an interpretation in terms of a closed-form and non-negative potential required by scalar auxiliary variable techniques. As a proof of concept, we generate synthetic data for the nonlinear transverse vibration of a string and show that the model can be trained to reproduce the nonlinear dynamics of the system. Sound examples are presented.

---

### Lunar-G2R: Geometry-to-Reflectance Learning for High-Fidelity Lunar BRDF Estimation

Authors: Clementine Grethen, Nicolas Menga, Roland Brochard, Geraldine Morin, Simone Gasparini, Jeremy Lebreton, Manuel Sanchez Gestido
Published: 2026-01-15
Categories: cs.CV
arXiv: https://arxiv.org/abs/2601.10449v1

We address the problem of estimating realistic, spatially varying reflectance for complex planetary surfaces such as the lunar regolith, which is critical for high-fidelity rendering and vision-based navigation. Existing lunar rendering pipelines rely on simplified or spatially uniform BRDF models whose parameters are difficult to estimate and fail to capture local reflectance variations, limiting photometric realism. We propose Lunar-G2R, a geometry-to-reflectance learning framework that predicts spatially varying BRDF parameters directly from a lunar digital elevation model (DEM), without requiring multi-view imagery, controlled illumination, or dedicated reflectance-capture hardware at inference time. The method leverages a U-Net trained with differentiable rendering to minimize photometric discrepancies between real orbital images and physically based renderings under known viewing and illumination geometry. Experiments on a geographically held-out region of the Tycho crater show that our approach reduces photometric error by 38 % compared to a state-of-the-art baseline, while achieving higher PSNR and SSIM and improved perceptual similarity, capturing fine-scale reflectance variations absent from spatially uniform models. To our knowledge, this is the first method to infer a spatially varying reflectance model directly from terrain geometry.

---

### Subjective evaluation of UHD video coded using VVC with LCEVC and ML-VVC

Authors: Naeem Ramzan, Muhammad Tufail Khan
Published: 2026-01-15
Categories: cs.MM, cs.CV
arXiv: https://arxiv.org/abs/2601.10448v1

This paper presents the results of a subjective quality assessment of a multilayer video coding configuration in which Low Complexity Enhancement Video Coding (LCEVC) is applied as an enhancement layer on top of a Versatile Video Coding (VVC) base layer. The evaluation follows the same test methodology and conditions previously defined for MPEG multilayer video coding assessments, with the LCEVC enhancement layer encoded using version 8.1 of the LCEVC Test Model (LTM). The test compares reconstructed UHD output generated from an HD VVC base layer with LCEVC enhancement against two reference cases: upsampled VVC base layer decoding and multilayer VVC (ML-VVC). Two operating points are considered, corresponding to enhancement layers representing approximately 10% and 50% of the total bitrate. Subjective assessment was conducted using the Degradation Category Rating (DCR) methodology with twenty five participants, across a dataset comprising fifteen SDR and HDR sequences. The reported results include Mean Opinion Scores (MOS) with associated 95% confidence intervals, enabling comparison of perceptual quality across coding approaches and operating points within the defined test scope.

---

### AgentGuardian: Learning Access Control Policies to Govern AI Agent Behavior

Authors: Nadya Abaev, Denis Klimov, Gerard Levinov, David Mimran, Yuval Elovici, Asaf Shabtai
Published: 2026-01-15
Categories: cs.CR, cs.AI, cs.LG
arXiv: https://arxiv.org/abs/2601.10440v1

Artificial intelligence (AI) agents are increasingly used in a variety of domains to automate tasks, interact with users, and make decisions based on data inputs. Ensuring that AI agents perform only authorized actions and handle inputs appropriately is essential for maintaining system integrity and preventing misuse. In this study, we introduce the AgentGuardian, a novel security framework that governs and protects AI agent operations by enforcing context-aware access-control policies. During a controlled staging phase, the framework monitors execution traces to learn legitimate agent behaviors and input patterns. From this phase, it derives adaptive policies that regulate tool calls made by the agent, guided by both real-time input context and the control flow dependencies of multi-step agent actions. Evaluation across two real-world AI agent applications demonstrates that AgentGuardian effectively detects malicious or misleading inputs while preserving normal agent functionality. Moreover, its control-flow-based governance mechanism mitigates hallucination-driven errors and other orchestration-level malfunctions.

---

### Development of Ontological Knowledge Bases by Leveraging Large Language Models

Authors: Le Ngoc Luyen, Marie-Hélène Abel, Philippe Gouspillou
Published: 2026-01-15
Categories: cs.IR, cs.AI
arXiv: https://arxiv.org/abs/2601.10436v2

Ontological Knowledge Bases (OKBs) play a vital role in structuring domain-specific knowledge and serve as a foundation for effective knowledge management systems. However, their traditional manual development poses significant challenges related to scalability, consistency, and adaptability. Recent advancements in Generative AI, particularly Large Language Models (LLMs), offer promising solutions for automating and enhancing OKB development. This paper introduces a structured, iterative methodology leveraging LLMs to optimize knowledge acquisition, automate ontology artifact generation, and enable continuous refinement cycles. We demonstrate this approach through a detailed case study focused on developing a user context profile ontology within the vehicle sales domain. Key contributions include significantly accelerated ontology construction processes, improved ontological consistency, effective bias mitigation, and enhanced transparency in the ontology engineering process. Our findings highlight the transformative potential of integrating LLMs into ontology development, notably improving scalability, integration capabilities, and overall efficiency in knowledge management systems.

---

### Are Language Models Models?

Authors: Philip Resnik
Published: 2026-01-15
Categories: cs.CL, cs.AI
arXiv: https://arxiv.org/abs/2601.10421v1

Futrell and Mahowald claim LMs "serve as model systems", but an assessment at each of Marr's three levels suggests the claim is clearly not true at the implementation level, poorly motivated at the algorithmic-representational level, and problematic at the computational theory level. LMs are good candidates as tools; calling them cognitive models overstates the case and unnecessarily feeds LLM hype.

---

### Reinforcement Learning with Multi-Step Lookahead Information Via Adaptive Batching

Authors: Nadav Merlis
Published: 2026-01-15
Categories: cs.LG, stat.ML
arXiv: https://arxiv.org/abs/2601.10418v1

We study tabular reinforcement learning problems with multiple steps of lookahead information. Before acting, the learner observes $\ell$ steps of future transition and reward realizations: the exact state the agent would reach and the rewards it would collect under any possible course of action. While it has been shown that such information can drastically boost the value, finding the optimal policy is NP-hard, and it is common to apply one of two tractable heuristics: processing the lookahead in chunks of predefined sizes ('fixed batching policies'), and model predictive control. We first illustrate the problems with these two approaches and propose utilizing the lookahead in adaptive (state-dependent) batches; we refer to such policies as adaptive batching policies (ABPs). We derive the optimal Bellman equations for these strategies and design an optimistic regret-minimizing algorithm that enables learning the optimal ABP when interacting with unknown environments. Our regret bounds are order-optimal up to a potential factor of the lookahead horizon $\ell$, which can usually be considered a small constant.

---

### LLMdoctor: Token-Level Flow-Guided Preference Optimization for Efficient Test-Time Alignment of Large Language Models

Authors: Tiesunlong Shen, Rui Mao, Jin Wang, Heming Sun, Jian Zhang, Xuejie Zhang, Erik Cambria
Published: 2026-01-15
Categories: cs.AI
arXiv: https://arxiv.org/abs/2601.10416v1

Aligning Large Language Models (LLMs) with human preferences is critical, yet traditional fine-tuning methods are computationally expensive and inflexible. While test-time alignment offers a promising alternative, existing approaches often rely on distorted trajectory-level signals or inefficient sampling, fundamentally capping performance and failing to preserve the generative diversity of the base model. This paper introduces LLMdoctor, a novel framework for efficient test-time alignment that operates via a patient-doctor paradigm. It integrates token-level reward acquisition with token-level flow-guided preference optimization (TFPO) to steer a large, frozen patient LLM with a smaller, specialized doctor model. Unlike conventional methods that rely on trajectory-level rewards, LLMdoctor first extracts fine-grained, token-level preference signals from the patient model's behavioral variations. These signals then guide the training of the doctor model via TFPO, which establishes flow consistency across all subtrajectories, enabling precise token-by-token alignment while inherently preserving generation diversity. Extensive experiments demonstrate that LLMdoctor significantly outperforms existing test-time alignment methods and even surpasses the performance of full fine-tuning approaches like DPO.

---

### LADFA: A Framework of Using Large Language Models and Retrieval-Augmented Generation for Personal Data Flow Analysis in Privacy Policies

Authors: Haiyue Yuan, Nikolay Matyunin, Ali Raza, Shujun Li
Published: 2026-01-15
Categories: cs.AI, cs.CR
arXiv: https://arxiv.org/abs/2601.10413v1

Privacy policies help inform people about organisations' personal data processing practices, covering different aspects such as data collection, data storage, and sharing of personal data with third parties. Privacy policies are often difficult for people to fully comprehend due to the lengthy and complex legal language used and inconsistent practices across different sectors and organisations. To help conduct automated and large-scale analyses of privacy policies, many researchers have studied applications of machine learning and natural language processing techniques, including large language models (LLMs). While a limited number of prior studies utilised LLMs for extracting personal data flows from privacy policies, our approach builds on this line of work by combining LLMs with retrieval-augmented generation (RAG) and a customised knowledge base derived from existing studies. This paper presents the development of LADFA, an end-to-end computational framework, which can process unstructured text in a given privacy policy, extract personal data flows and construct a personal data flow graph, and conduct analysis of the data flow graph to facilitate insight discovery. The framework consists of a pre-processor, an LLM-based processor, and a data flow post-processor. We demonstrated and validated the effectiveness and accuracy of the proposed approach by conducting a case study that involved examining ten selected privacy policies from the automotive industry. Moreover, it is worth noting that LADFA is designed to be flexible and customisable, making it suitable for a range of text-based analysis tasks beyond privacy policy analysis.

---

### TF3-RO-50M: Training Compact Romanian Language Models from Scratch on Synthetic Moral Microfiction

Authors: Mihai Dan Nadas, Laura Diosan, Andreea Tomescu, Andrei Piscoran
Published: 2026-01-15
Categories: cs.CL
arXiv: https://arxiv.org/abs/2601.10410v1

Recent advances in synthetic data generation have shown that compact language models can be trained effectively when the underlying corpus is structurally controlled and linguistically coherent. However, for morphologically rich and computationally under-resourced languages such as Romanian, there is still no openly documented, end-to-end pipeline that unifies tokenizer design, preprocessing, pretraining, compression, evaluation, and large-scale synthetic data generation in a reproducible framework. Building on TF1, a three-million-story English fable dataset, and TF2, which extends TF1 through high-quality Romanian translations, we introduce TF3-RO, a Romanian-centric language modeling pipeline spanning tokenizer training, from-scratch model development, and Romanian-native dataset generation. TF3-RO constructs Romanian-specific BPE and Unigram tokenizers from a linguistically informed corpus to mitigate token inflation induced by Romanian morphology. Using long-sequence packed training, we pretrain a 51.65M-parameter LLaMA-style Transformer entirely from scratch. The model is subsequently optimized through quantization, structured pruning, and logit-based knowledge distillation, yielding a compact 26.45M-parameter student model with tied embeddings and strong deployment characteristics. Using this distilled model, TF3-RO generates three million Romanian-native synthetic fables via a controlled combinatorial prompting framework. Across all stages, the pipeline integrates a comprehensive evaluation suite combining intrinsic metrics, Romanian agreement probes, entity coherence, rule-based grammar checking, and LLM-based assessment. TF3-RO provides a reproducible and linguistically grounded framework for training compact Romanian language models and producing large-scale synthetic narrative corpora.

---

### CS-GBA: A Critical Sample-based Gradient-guided Backdoor Attack for Offline Reinforcement Learning

Authors: Yuanjie Zhao, Junnan Qiu, Yue Ding, Jie Li
Published: 2026-01-15
Categories: cs.LG
arXiv: https://arxiv.org/abs/2601.10407v1

Offline Reinforcement Learning (RL) enables policy optimization from static datasets but is inherently vulnerable to backdoor attacks. Existing attack strategies typically struggle against safety-constrained algorithms (e.g., CQL) due to inefficient random poisoning and the use of easily detectable Out-of-Distribution (OOD) triggers. In this paper, we propose CS-GBA (Critical Sample-based Gradient-guided Backdoor Attack), a novel framework designed to achieve high stealthiness and destructiveness under a strict budget. Leveraging the theoretical insight that samples with high Temporal Difference (TD) errors are pivotal for value function convergence, we introduce an adaptive Critical Sample Selection strategy that concentrates the attack budget on the most influential transitions. To evade OOD detection, we propose a Correlation-Breaking Trigger mechanism that exploits the physical mutual exclusivity of state features (e.g., 95th percentile boundaries) to remain statistically concealed. Furthermore, we replace the conventional label inversion with a Gradient-Guided Action Generation mechanism, which searches for worst-case actions within the data manifold using the victim Q-network's gradient. Empirical results on D4RL benchmarks demonstrate that our method significantly outperforms state-of-the-art baselines, achieving high attack success rates against representative safety-constrained algorithms with a minimal 5% poisoning budget, while maintaining the agent's performance in clean environments.

---

### ErrEval: Error-Aware Evaluation for Question Generation through Explicit Diagnostics

Authors: Weiping Fu, Bifan Wei, Jingyi Hao, Yushun Zhang, Jian Zhang, Jiaxin Wang, Bo Li, Yu He, Lingling Zhang, Jun Liu
Published: 2026-01-15
Categories: cs.AI
arXiv: https://arxiv.org/abs/2601.10406v1

Automatic Question Generation (QG) often produces outputs with critical defects, such as factual hallucinations and answer mismatches. However, existing evaluation methods, including LLM-based evaluators, mainly adopt a black-box and holistic paradigm without explicit error modeling, leading to the neglect of such defects and overestimation of question quality. To address this issue, we propose ErrEval, a flexible and Error-aware Evaluation framework that enhances QG evaluation through explicit error diagnostics. Specifically, ErrEval reformulates evaluation as a two-stage process of error diagnosis followed by informed scoring. At the first stage, a lightweight plug-and-play Error Identifier detects and categorizes common errors across structural, linguistic, and content-related aspects. These diagnostic signals are then incorporated as explicit evidence to guide LLM evaluators toward more fine-grained and grounded judgments. Extensive experiments on three benchmarks demonstrate the effectiveness of ErrEval, showing that incorporating explicit diagnostics improves alignment with human judgments. Further analyses confirm that ErrEval effectively mitigates the overestimation of low-quality questions.

---

### Discrete Feynman-Kac Correctors

Authors: Mohsin Hasan, Viktor Ohanesian, Artem Gazizov, Yoshua Bengio, Alán Aspuru-Guzik, Roberto Bondesan, Marta Skreta, Kirill Neklyudov
Published: 2026-01-15
Categories: cs.LG
arXiv: https://arxiv.org/abs/2601.10403v1

Discrete diffusion models have recently emerged as a promising alternative to the autoregressive approach for generating discrete sequences. Sample generation via gradual denoising or demasking processes allows them to capture hierarchical non-sequential interdependencies in the data. These custom processes, however, do not assume a flexible control over the distribution of generated samples. We propose Discrete Feynman-Kac Correctors, a framework that allows for controlling the generated distribution of discrete masked diffusion models at inference time. We derive Sequential Monte Carlo (SMC) algorithms that, given a trained discrete diffusion model, control the temperature of the sampled distribution (i.e. perform annealing), sample from the product of marginals of several diffusion processes (e.g. differently conditioned processes), and sample from the product of the marginal with an external reward function, producing likely samples from the target distribution that also have high reward. Notably, our framework does not require any training of additional models or fine-tuning of the original model. We illustrate the utility of our framework in several applications including: efficient sampling from the annealed Boltzmann distribution of the Ising model, improving the performance of language models for code generation and amortized learning, as well as reward-tilted protein sequence generation.

---

### Toward Ultra-Long-Horizon Agentic Science: Cognitive Accumulation for Machine Learning Engineering

Authors: Xinyu Zhu, Yuzhu Cai, Zexi Liu, Bingyang Zheng, Cheng Wang, Rui Ye, Yuzhi Zhang, Linfeng Zhang, Weinan E, Siheng Chen, Yanfeng Wang
Published: 2026-01-15
Categories: cs.AI
arXiv: https://arxiv.org/abs/2601.10402v5

The advancement of artificial intelligence toward agentic science is currently bottlenecked by the challenge of ultra-long-horizon autonomy, the ability to sustain strategic coherence and iterative correction over experimental cycles spanning days or weeks. While Large Language Models (LLMs) have demonstrated prowess in short-horizon reasoning, they are easily overwhelmed by execution details in the high-dimensional, delayed-feedback environments of real-world research, failing to consolidate sparse feedback into coherent long-term guidance. Here, we present ML-Master 2.0, an autonomous agent that masters ultra-long-horizon machine learning engineering (MLE) which is a representative microcosm of scientific discovery. By reframing context management as a process of cognitive accumulation, our approach introduces Hierarchical Cognitive Caching (HCC), a multi-tiered architecture inspired by computer systems that enables the structural differentiation of experience over time. By dynamically distilling transient execution traces into stable knowledge and cross-task wisdom, HCC allows agents to decouple immediate execution from long-term experimental strategy, effectively overcoming the scaling limits of static context windows. In evaluations on OpenAI's MLE-Bench under 24-hour budgets, ML-Master 2.0 achieves a state-of-the-art medal rate of 56.44%. Our findings demonstrate that ultra-long-horizon autonomy provides a scalable blueprint for AI capable of autonomous exploration beyond human-precedent complexities.

---

### LatentRefusal: Latent-Signal Refusal for Unanswerable Text-to-SQL Queries

Authors: Xuancheng Ren, Shijing Hu, Zhihui Lu, Jiangqi Huang, Qiang Duan
Published: 2026-01-15
Categories: cs.AI
arXiv: https://arxiv.org/abs/2601.10398v3

In LLM-based text-to-SQL systems, unanswerable and underspecified user queries may generate not only incorrect text but also executable programs that yield misleading results or violate safety constraints, posing a major barrier to safe deployment. Existing refusal strategies for such queries either rely on output-level instruction following, which is brittle due to model hallucinations, or estimate output uncertainty, which adds complexity and overhead. To address this challenge, we formalize safe refusal in text-to-SQL systems as an answerability-gating problem and propose LatentRefusal, a latent-signal refusal mechanism that predicts query answerability from intermediate hidden activations of a large language model. We introduce the Tri-Residual Gated Encoder, a lightweight probing architecture, to suppress schema noise and amplify sparse, localized cues of question-schema mismatch that indicate unanswerability. Extensive empirical evaluations across diverse ambiguous and unanswerable settings, together with ablation studies and interpretability analyses, demonstrate the effectiveness of the proposed approach and show that LatentRefusal provides an attachable and efficient safety layer for text-to-SQL systems. Across four benchmarks, LatentRefusal improves average F1 to 88.5 percent on both backbones while adding approximately 2 milliseconds of probe overhead.

---

### Multi-Temporal Frames Projection for Dynamic Processes Fusion in Fluorescence Microscopy

Authors: Hassan Eshkiki, Sarah Costa, Mostafa Mohammadpour, Farinaz Tanhaei, Christopher H. George, Fabio Caraffini
Published: 2026-01-15
Categories: cs.CV
arXiv: https://arxiv.org/abs/2601.10392v1

Fluorescence microscopy is widely employed for the analysis of living biological samples; however, the utility of the resulting recordings is frequently constrained by noise, temporal variability, and inconsistent visualisation of signals that oscillate over time. We present a unique computational framework that integrates information from multiple time-resolved frames into a single high-quality image, while preserving the underlying biological content of the original video. We evaluate the proposed method through an extensive number of configurations (n = 111) and on a challenging dataset comprising dynamic, heterogeneous, and morphologically complex 2D monolayers of cardiac cells. Results show that our framework, which consists of a combination of explainable techniques from different computer vision application fields, is capable of generating composite images that preserve and enhance the quality and information of individual microscopy frames, yielding 44% average increase in cell count compared to previous methods. The proposed pipeline is applicable to other imaging domains that require the fusion of multi-temporal image stacks into high-quality 2D images, thereby facilitating annotation and downstream segmentation.

---

### INDIC DIALECT: A Multi Task Benchmark to Evaluate and Translate in Indian Language Dialects

Authors: Tarun Sharma, Manikandan Ravikiran, Sourava Kumar Behera, Pramit Bhattacharya, Arnab Bhattacharya, Rohit Saluja
Published: 2026-01-15
Categories: cs.CL
arXiv: https://arxiv.org/abs/2601.10388v1

Recent NLP advances focus primarily on standardized languages, leaving most low-resource dialects under-served especially in Indian scenarios. In India, the issue is particularly important: despite Hindi being the third most spoken language globally (over 600 million speakers), its numerous dialects remain underrepresented. The situation is similar for Odia, which has around 45 million speakers. While some datasets exist which contain standard Hindi and Odia languages, their regional dialects have almost no web presence. We introduce INDIC-DIALECT, a human-curated parallel corpus of 13k sentence pairs spanning 11 dialects and 2 languages: Hindi and Odia. Using this corpus, we construct a multi-task benchmark with three tasks: dialect classification, multiple-choice question (MCQ) answering, and machine translation (MT). Our experiments show that LLMs like GPT-4o and Gemini 2.5 perform poorly on the classification task. While fine-tuned transformer based models pretrained on Indian languages substantially improve performance e.g., improving F1 from 19.6\% to 89.8\% on dialect classification. For dialect to language translation, we find that hybrid AI model achieves highest BLEU score of 61.32 compared to the baseline score of 23.36. Interestingly, due to complexity in generating dialect sentences, we observe that for language to dialect translation the ``rule-based followed by AI" approach achieves best BLEU score of 48.44 compared to the baseline score of 27.59. INDIC-DIALECT thus is a new benchmark for dialect-aware Indic NLP, and we plan to release it as open source to support further work on low-resource Indian dialects.

---

### The Assistant Axis: Situating and Stabilizing the Default Persona of Language Models

Authors: Christina Lu, Jack Gallagher, Jonathan Michala, Kyle Fish, Jack Lindsey
Published: 2026-01-15
Categories: cs.CL
arXiv: https://arxiv.org/abs/2601.10387v1

Large language models can represent a variety of personas but typically default to a helpful Assistant identity cultivated during post-training. We investigate the structure of the space of model personas by extracting activation directions corresponding to diverse character archetypes. Across several different models, we find that the leading component of this persona space is an "Assistant Axis," which captures the extent to which a model is operating in its default Assistant mode. Steering towards the Assistant direction reinforces helpful and harmless behavior; steering away increases the model's tendency to identify as other entities. Moreover, steering away with more extreme values often induces a mystical, theatrical speaking style. We find this axis is also present in pre-trained models, where it primarily promotes helpful human archetypes like consultants and coaches and inhibits spiritual ones. Measuring deviations along the Assistant Axis predicts "persona drift," a phenomenon where models slip into exhibiting harmful or bizarre behaviors that are uncharacteristic of their typical persona. We find that persona drift is often driven by conversations demanding meta-reflection on the model's processes or featuring emotionally vulnerable users. We show that restricting activations to a fixed region along the Assistant Axis can stabilize model behavior in these scenarios -- and also in the face of adversarial persona-based jailbreaks. Our results suggest that post-training steers models toward a particular region of persona space but only loosely tethers them to it, motivating work on training and steering strategies that more deeply anchor models to a coherent persona.

---