# Research Papers — 2026-01 · bundle page 25 of 189

> Full-text markdown of papers #1201–1250 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/25
[← Previous bundle](https://aipod.dev/papers/archive/2026-01/page/24.md)
[Next bundle →](https://aipod.dev/papers/archive/2026-01/page/26.md)

### KID: Knowledge-Injected Dual-Head Learning for Knowledge-Grounded Harmful Meme Detection

Authors: Yaocong Li, Leihan Zhang, Le Zhang, Qiang Yan
Published: 2026-01-29
Categories: cs.CL, cs.AI
arXiv: https://arxiv.org/abs/2601.21796v1

Internet memes have become pervasive carriers of digital culture on social platforms. However, their heavy reliance on metaphors and sociocultural context also makes them subtle vehicles for harmful content, posing significant challenges for automated content moderation. Existing approaches primarily focus on intra-modal and inter-modal signal analysis, while the understanding of implicit toxicity often depends on background knowledge that is not explicitly present in the meme itself. To address this challenge, we propose KID, a Knowledge-Injected Dual-Head Learning framework for knowledge-grounded harmful meme detection. KID adopts a label-constrained distillation paradigm to decompose complex meme understanding into structured reasoning chains that explicitly link visual evidence, background knowledge, and classification labels. These chains guide the learning process by grounding external knowledge in meme-specific contexts. In addition, KID employs a dual-head architecture that jointly optimizes semantic generation and classification objectives, enabling aligned linguistic reasoning while maintaining stable decision boundaries. Extensive experiments on five multilingual datasets spanning English, Chinese, and low-resource Bengali demonstrate that KID achieves SOTA performance on both binary and multi-label harmful meme detection tasks, improving over previous best methods by 2.1%--19.7% across primary evaluation metrics. Ablation studies further confirm the effectiveness of knowledge injection and dual-head joint learning, highlighting their complementary contributions to robust and generalizable meme understanding. The code and data are available at https://github.com/PotatoDog1669/KID.

---

### Effective LoRA Adapter Routing using Task Representations

Authors: Akash Dhasade, Anne-Marie Kermarrec, Igor Pavlovic, Diana Petrescu, Rafael Pires, Mathis Randl, Martijn de Vos
Published: 2026-01-29
Categories: cs.LG, cs.AI
arXiv: https://arxiv.org/abs/2601.21795v2

Low-rank adaptation (LoRA) enables parameter efficient specialization of large language models (LLMs) through modular adapters, resulting in rapidly growing public adapter pools spanning diverse tasks. Effectively using these adapters requires routing: selecting and composing the appropriate adapters for a query. We introduce LORAUTER, a novel routing framework that selects and composes LoRA adapters using task representations rather than adapter characteristics. Unlike existing approaches that map queries directly to adapters, LORAUTER routes queries via task embeddings derived from small validation sets and does not require adapter training data. By operating at the task level, LORAUTER achieves efficient routing that scales with the number of tasks rather than the number of adapters. Experiments across multiple tasks show that LORAUTER consistently outperforms baseline routing approaches, matching Oracle performance (101.2%) when task-aligned adapters exist and achieving state-of-the-art results on unseen tasks (+5.2 points). We further demonstrate the robustness of LORAUTER to very large, noisy adapter pools by scaling it to over 1500 adapters.

---

### Knowledge Vector Weakening: Efficient Training-free Unlearning for Large Vision-Language Models

Authors: Yejin Kim, Dongjun Hwang, Sungmin Cha, Junsuk Choe
Published: 2026-01-29
Categories: cs.LG
arXiv: https://arxiv.org/abs/2601.21794v1

Large Vision-Language Models (LVLMs) are widely adopted for their strong multimodal capabilities, yet they raise serious concerns such as privacy leakage and harmful content generation. Machine unlearning has emerged as a promising solution for removing the influence of specific data from trained models. However, existing approaches largely rely on gradient-based optimization, incurring substantial computational costs for large-scale LVLMs. To address this limitation, we propose Knowledge Vector Weakening (KVW), a training-free unlearning method that directly intervenes in the full model without gradient computation. KVW identifies knowledge vectors that are activated during the model's output generation on the forget set and progressively weakens their contributions, thereby preventing the model from exploiting undesirable knowledge. Experiments on the MLLMU and CLEAR benchmarks demonstrate that KVW achieves a stable forget-retain trade-off while significantly improving computational efficiency over gradient-based and LoRA-based unlearning methods.

---

### NetMamba+: A Framework of Pre-trained Models for Efficient and Accurate Network Traffic Classification

Authors: Tongze Wang, Xiaohui Xie, Wenduo Wang, Chuyi Wang, Jinzhou Liu, Boyan Huang, Yannan Hu, Youjian Zhao, Yong Cui
Published: 2026-01-29
Categories: cs.LG
arXiv: https://arxiv.org/abs/2601.21792v1

With the rapid growth of encrypted network traffic, effective traffic classification has become essential for network security and quality of service management. Current machine learning and deep learning approaches for traffic classification face three critical challenges: computational inefficiency of Transformer architectures, inadequate traffic representations with loss of crucial byte-level features while retaining detrimental biases, and poor handling of long-tail distributions in real-world data. We propose NetMamba+, a framework that addresses these challenges through three key innovations: (1) an efficient architecture considering Mamba and Flash Attention mechanisms, (2) a multimodal traffic representation scheme that preserves essential traffic information while eliminating biases, and (3) a label distribution-aware fine-tuning strategy. Evaluation experiments on massive datasets encompassing four main classification tasks showcase NetMamba+'s superior classification performance compared to state-of-the-art baselines, with improvements of up to 6.44\% in F1 score. Moreover, NetMamba+ demonstrates excellent efficiency, achieving 1.7x higher inference throughput than the best baseline while maintaining comparably low memory usage. Furthermore, NetMamba+ exhibits superior few-shot learning abilities, achieving better classification performance with fewer labeled data. Additionally, we implement an online traffic classification system that demonstrates robust real-world performance with a throughput of 261.87 Mb/s. As the first framework to adapt Mamba architecture for network traffic classification, NetMamba+ opens new possibilities for efficient and accurate traffic analysis in complex network environments.

---

### ECSEL: Explainable Classification via Signomial Equation Learning

Authors: Adia Lumadjeng, Ilker Birbil, Erman Acar
Published: 2026-01-29
Categories: cs.LG, cs.AI, stat.ML
arXiv: https://arxiv.org/abs/2601.21789v2

We introduce ECSEL, an explainable classification method that learns formal expressions in the form of signomial equations, motivated by the observation that many symbolic regression benchmarks admit compact signomial structure. ECSEL directly constructs a structural, closed-form expression that serves as both a classifier and an explanation. On standard symbolic regression benchmarks, our method recovers a larger fraction of target equations than competing state-of-the-art approaches while requiring substantially less computation. Leveraging this efficiency, ECSEL achieves classification accuracy competitive with established machine learning models without sacrificing interpretability. Further, we show that ECSEL satisfies some desirable properties regarding global feature behavior, decision-boundary analysis, and local feature attributions. Experiments on benchmark datasets and two real-world case studies i.e., e-commerce and fraud detection, demonstrate that the learned equations expose dataset biases, support counterfactual reasoning, and yield actionable insights.

---

### Assessing the Business Process Modeling Competences of Large Language Models

Authors: Chantale Lauer, Peter Pfeiffer, Alexander Rombach, Nijat Mehdiyev
Published: 2026-01-29
Categories: cs.SE, cs.AI
arXiv: https://arxiv.org/abs/2601.21787v2

The creation of Business Process Model and Notation (BPMN) models is a complex and time-consuming task requiring both domain knowledge and proficiency in modeling conventions. Recent advances in large language models (LLMs) have significantly expanded the possibilities for generating BPMN models directly from natural language, building upon earlier text-to-process methods with enhanced capabilities in handling complex descriptions. However, there is a lack of systematic evaluations of LLM-generated process models. Current efforts either use LLM-as-a-judge approaches or do not consider established dimensions of model quality. To this end, we introduce BEF4LLM, a novel LLM evaluation framework comprising four perspectives: syntactic quality, pragmatic quality, semantic quality, and validity. Using BEF4LLM, we conduct a comprehensive analysis of open-source LLMs and benchmark their performance against human modeling experts. Results indicate that LLMs excel in syntactic and pragmatic quality, while humans outperform LLMs in semantic aspects; however, the differences in scores are relatively modest, highlighting LLMs' competitive potential despite challenges in validity and semantic quality. The insights highlight current strengths and limitations of using LLMs for BPMN modeling and guide future model development and fine-tuning. Addressing these areas is essential for advancing the practical deployment of LLMs in business process modeling.

---

### Synthetic-to-Real Domain Bridging for Single-View 3D Reconstruction of Ships for Maritime Monitoring

Authors: Borja Carrillo-Perez, Felix Sattler, Angel Bueno Rodriguez, Maurice Stephan, Sarah Barnes
Published: 2026-01-29
Categories: cs.CV, cs.AI, cs.GR
arXiv: https://arxiv.org/abs/2601.21786v1

Three-dimensional (3D) reconstruction of ships is an important part of maritime monitoring, allowing improved visualization, inspection, and decision-making in real-world monitoring environments. However, most state-ofthe-art 3D reconstruction methods require multi-view supervision, annotated 3D ground truth, or are computationally intensive, making them impractical for real-time maritime deployment. In this work, we present an efficient pipeline for single-view 3D reconstruction of real ships by training entirely on synthetic data and requiring only a single view at inference. Our approach uses the Splatter Image network, which represents objects as sparse sets of 3D Gaussians for rapid and accurate reconstruction from single images. The model is first fine-tuned on synthetic ShapeNet vessels and further refined with a diverse custom dataset of 3D ships, bridging the domain gap between synthetic and real-world imagery. We integrate a state-of-the-art segmentation module based on YOLOv8 and custom preprocessing to ensure compatibility with the reconstruction network. Postprocessing steps include real-world scaling, centering, and orientation alignment, followed by georeferenced placement on an interactive web map using AIS metadata and homography-based mapping. Quantitative evaluation on synthetic validation data demonstrates strong reconstruction fidelity, while qualitative results on real maritime images from the ShipSG dataset confirm the potential for transfer to operational maritime settings. The final system provides interactive 3D inspection of real ships without requiring real-world 3D annotations. This pipeline provides an efficient, scalable solution for maritime monitoring and highlights a path toward real-time 3D ship visualization in practical applications. Interactive demo: https://dlr-mi.github.io/ship3d-demo/.

---

### Quantum LEGO Learning: A Modular Design Principle for Hybrid Artificial Intelligence

Authors: Jun Qi, Chao-Han Huck Yang, Pin-Yu Chen, Min-Hsiu Hsieh, Hector Zenil, Jesper Tegner
Published: 2026-01-29
Categories: cs.LG, quant-ph
arXiv: https://arxiv.org/abs/2601.21780v1

Hybrid quantum-classical learning models increasingly integrate neural networks with variational quantum circuits (VQCs) to exploit complementary inductive biases. However, many existing approaches rely on tightly coupled architectures or task-specific encoders, limiting conceptual clarity, generality, and transferability across learning settings. In this work, we introduce Quantum LEGO Learning, a modular and architecture-agnostic learning framework that treats classical and quantum components as reusable, composable learning blocks with well-defined roles. Within this framework, a pre-trained classical neural network serves as a frozen feature block, while a VQC acts as a trainable adaptive module that operates on structured representations rather than raw inputs. This separation enables efficient learning under constrained quantum resources and provides a principled abstraction for analyzing hybrid models. We develop a block-wise generalization theory that decomposes learning error into approximation and estimation components, explicitly characterizing how the complexity and training status of each block influence overall performance. Our analysis generalizes prior tensor-network-specific results and identifies conditions under which quantum modules provide representational advantages over comparably sized classical heads. Empirically, we validate the framework through systematic block-swap experiments across frozen feature extractors and both quantum and classical adaptive heads. Experiments on quantum dot classification demonstrate stable optimization, reduced sensitivity to qubit count, and robustness to realistic noise.

---

### Error Amplification Limits ANN-to-SNN Conversion in Continuous Control

Authors: Zijie Xu, Zihan Huang, Yiting Dong, Kang Chen, Wenxuan Liu, Zhaofei Yu
Published: 2026-01-29
Categories: cs.NE, cs.LG
arXiv: https://arxiv.org/abs/2601.21778v3

Spiking Neural Networks (SNNs) can achieve competitive performance by converting already existing well-trained Artificial Neural Networks (ANNs), avoiding further costly training. This property is particularly attractive in Reinforcement Learning (RL), where training through environment interaction is expensive and potentially unsafe. However, existing conversion methods perform poorly in continuous control, where suitable baselines are largely absent. We identify error amplification as the key cause: small action approximation errors become temporally correlated across decision steps, inducing cumulative state distribution shift and severe performance degradation. To address this issue, we propose Cross-Step Residual Potential Initialization (CRPI), a lightweight gradient-free mechanism that carries over residual membrane potentials across decision steps to suppress temporally correlated errors. Experiments on continuous control benchmarks with both vector and visual observations demonstrate that CRPI can be integrated into existing conversion pipelines and substantially recovers lost performance. Our results highlight continuous control as a critical and challenging benchmark for ANN-to-SNN conversion, where small errors can be strongly amplified and impact performance. Code is available at https://github.com/xuzijie32/ANN2SNN-CRPI.

---

### Differentiable Knapsack and Top-k Operators via Dynamic Programming

Authors: Germain Vivier-Ardisson, Michaël E. Sander, Axel Parmentier, Mathieu Blondel
Published: 2026-01-29
Categories: cs.LG
arXiv: https://arxiv.org/abs/2601.21775v1

Knapsack and Top-k operators are useful for selecting discrete subsets of variables. However, their integration into neural networks is challenging as they are piecewise constant, yielding gradients that are zero almost everywhere. In this paper, we propose a unified framework casting these operators as dynamic programs, and derive differentiable relaxations by smoothing the underlying recursions. On the algorithmic side, we develop efficient parallel algorithms supporting both deterministic and stochastic forward passes, and vector-Jacobian products for the backward pass. On the theoretical side, we prove that Shannon entropy is the unique regularization choice yielding permutation-equivariant operators, and characterize regularizers inducing sparse selections. Finally, on the experimental side, we demonstrate our framework on a decision-focused learning benchmark, a constrained dynamic assortment RL problem, and an extension of discrete VAEs.

---

### Abstract Concept Modelling in Conceptual Spaces: A Study on Chess Strategies

Authors: Hadi Banaee, Stephanie Lowry
Published: 2026-01-29
Categories: cs.AI
arXiv: https://arxiv.org/abs/2601.21771v1

We present a conceptual space framework for modelling abstract concepts that unfold over time, demonstrated through a chess-based proof-of-concept. Strategy concepts, such as attack or sacrifice, are represented as geometric regions across interpretable quality dimensions, with chess games instantiated and analysed as trajectories whose directional movement toward regions enables recognition of intended strategies. This approach also supports dual-perspective modelling, capturing how players interpret identical situations differently. Our implementation demonstrates the feasibility of trajectory-based concept recognition, with movement patterns aligning with expert commentary. This work explores extending the conceptual spaces theory to temporally realised, goal-directed concepts. The approach establishes a foundation for broader applications involving sequential decision-making and supports integration with knowledge evolution mechanisms for learning and refining abstract concepts over time.

---

### Zonkey: A Hierarchical Diffusion Language Model with Differentiable Tokenization and Probabilistic Attention

Authors: Alon Rozental
Published: 2026-01-29
Categories: cs.CL
arXiv: https://arxiv.org/abs/2601.21768v1

Large language models (LLMs) have revolutionized natural language processing, yet they remain constrained by fixed, non-differentiable tokenizers like Byte Pair Encoding (BPE), which hinder end-to-end optimization and adaptability to noisy or domain-specific data. We introduce Zonkey, a hierarchical diffusion model that addresses these limitations through a fully trainable pipeline from raw characters to document-level representations. At its core is a differentiable tokenizer (Segment Splitter) that learns probabilistic beginning-of-sequence (BOS) decisions, enabling adaptive splits that emerge as linguistically meaningful (e.g., word boundaries at spaces, sentence starts at periods) without explicit supervision. This differentiability is enabled by our novel Probabilistic Attention mechanism, which incorporates position-specific existence probabilities to simulate soft masking over theoretically infinite sequences while preserving gradients. Sequences decay probabilistically rather than relying on end-of-sequence tokens, supporting variable-length outputs. Hierarchical levels compress sequences into higher abstractions (e.g., character n-grams to word-like vectors, then sentence-like), with reconstruction via our Denoising Diffusion Mixed Model (DDMM) for stable and efficient denoising in latent space. A Stitcher ensures overlap invariance across segments. Trained end-to-end on Wikipedia, Zonkey generates coherent, variable-length text from noise, demonstrating emergent hierarchies and promising qualitative alignment to data distributions compared to entropy-based learnable tokenizers. Our approach advances toward fully gradient-based LLMs, with potential for better domain adaptation and scalable generation. We release the source code for training and reproducing our experiments.

---

### Evaluating ChatGPT on Medical Information Extraction Tasks: Performance, Explainability and Beyond

Authors: Liz Li, Wei Zhu
Published: 2026-01-29
Categories: cs.CL
arXiv: https://arxiv.org/abs/2601.21767v2

Large Language Models (LLMs) like ChatGPT have demonstrated amazing capabilities in comprehending user intents and generate reasonable and useful responses. Beside their ability to chat, their capabilities in various natural language processing (NLP) tasks are of interest to the research community. In this paper, we focus on assessing the overall ability of ChatGPT in 4 different medical information extraction (MedIE) tasks across 6 benchmark datasets. We present the systematically analysis by measuring ChatGPT's performance, explainability, confidence, faithfulness, and uncertainty. Our experiments reveal that: (a) ChatGPT's performance scores on MedIE tasks fall behind those of the fine-tuned baseline models. (b) ChatGPT can provide high-quality explanations for its decisions, however, ChatGPT is over-confident in its predcitions. (c) ChatGPT demonstrates a high level of faithfulness to the original text in the majority of cases. (d) The uncertainty in generation causes uncertainty in information extraction results, thus may hinder its applications in MedIE tasks.

---

### CoFrGeNet: Continued Fraction Architectures for Language Generation

Authors: Amit Dhurandhar, Vijil Chenthamarakshan, Dennis Wei, Tejaswini Pedapati, Karthikeyan Natesan Ramamurthy, Rahul Nair
Published: 2026-01-29
Categories: cs.CL, cs.AI
arXiv: https://arxiv.org/abs/2601.21766v5

Transformers are arguably the preferred architecture for language generation. In this paper, inspired by continued fractions, we introduce a new function class for generative modeling. The architecture family implementing this function class is named CoFrGeNets - Continued Fraction Generative Networks. We design novel architectural components based on this function class that can replace Multi-head Attention and Feed-Forward Networks in Transformer blocks while requiring much fewer parameters. We derive custom gradient formulations to optimize the proposed components more accurately and efficiently than using standard PyTorch-based gradients. Our components are a plug-in replacement requiring little change in training or inference procedures that have already been put in place for Transformer-based models thus making our approach easy to incorporate in large industrial workflows. We experiment on two very different transformer architectures GPT2-xl (1.5B) and Llama3 (3.2B), where the former we pre-train on OpenWebText and GneissWeb, while the latter we pre-train on the docling data mix which consists of nine different datasets. Results show that the performance on downstream classification, Q\& A, reasoning and text understanding tasks of our models is competitive and sometimes even superior to the original models with $\frac{2}{3}$ to $\frac{1}{2}$ the parameters and shorter pre-training time. We believe that future implementations customized to hardware will further bring out the true potential of our architectures.

---

### Mean-field Variational Bayes for Sparse Probit Regression

Authors: Augusto Fasano, Giovanni Rebaudo
Published: 2026-01-29
Categories: stat.CO, stat.ME, stat.ML
arXiv: https://arxiv.org/abs/2601.21765v2

We consider Bayesian variable selection for binary outcomes under a probit link with a spike-and-slab prior on the regression coefficients. Motivated by the computational challenges encountered by Markov chain Monte Carlo (MCMC) samplers in high-dimensional regimes, we develop a mean-field variational Bayes approximation in which all variational factors admit closed-form updates, and the evidence lower bound is available in closed form. This, in turn, allows the development of an efficient coordinate ascent variational inference algorithm to find the optimal values of the variational parameters. The approach produces posterior inclusion probabilities and parameter estimates, enabling interpretable selection and prediction within a single framework. As shown in both simulated and real data applications, the proposed method successfully identifies the important variables and is orders of magnitude faster than MCMC, while maintaining comparable accuracy.

---

### Zero-Shot Statistical Downscaling via Diffusion Posterior Sampling

Authors: Ruian Tie, Wenbo Xiong, Zhengyu Shi, Xinyu Su, Chenyu jiang, Libo Wu, Hao Li
Published: 2026-01-29
Categories: cs.AI
arXiv: https://arxiv.org/abs/2601.21760v2

Conventional supervised climate downscaling struggles to generalize to Global Climate Models (GCMs) due to the lack of paired training data and inherent domain gaps relative to reanalysis. Meanwhile, current zero-shot methods suffer from physical inconsistencies and vanishing gradient issues under large scaling factors. We propose Zero-Shot Statistical Downscaling (ZSSD), a zero-shot framework that performs statistical downscaling without paired data during training. ZSSD leverages a Physics-Consistent Climate Prior learned from reanalysis data, conditioned on geophysical boundaries and temporal information to enforce physical validity. Furthermore, to enable robust inference across varying GCMs, we introduce Unified Coordinate Guidance. This strategy addresses the vanishing gradient problem in vanilla DPS and ensures consistency with large-scale fields. Results show that ZSSD significantly outperforms existing zero-shot baselines in 99th percentile errors and successfully reconstructs complex weather events, such as tropical cyclones, across heterogeneous GCMs.

---

### Influence Guided Sampling for Domain Adaptation of Text Retrievers

Authors: Meet Doshi, Vishwajeet Kumar, Yulong Li, Jaydeep Sen
Published: 2026-01-29
Categories: cs.IR, cs.CL
arXiv: https://arxiv.org/abs/2601.21759v1

General-purpose open-domain dense retrieval systems are usually trained with a large, eclectic mix of corpora and search tasks. How should these diverse corpora and tasks be sampled for training? Conventional approaches sample them uniformly, proportional to their instance population sizes, or depend on human-level expert supervision. It is well known that the training data sampling strategy can greatly impact model performance. However, how to find the optimal strategy has not been adequately studied in the context of embedding models. We propose Inf-DDS, a novel reinforcement learning driven sampling framework that adaptively reweighs training datasets guided by influence-based reward signals and is much more lightweight with respect to GPU consumption. Our technique iteratively refines the sampling policy, prioritizing datasets that maximize model performance on a target development set. We evaluate the efficacy of our sampling strategy on a wide range of text retrieval tasks, demonstrating strong improvements in retrieval performance and better adaptation compared to existing gradient-based sampling methods, while also being 1.5x to 4x cheaper in GPU compute. Our sampling strategy achieves a 5.03 absolute NDCG@10 improvement while training a multilingual bge-m3 model and an absolute NDCG@10 improvement of 0.94 while training all-MiniLM-L6-v2, even when starting from expert-assigned weights on a large pool of training datasets.

---

### EWSJF: An Adaptive Scheduler with Hybrid Partitioning for Mixed-Workload LLM Inference

Authors: Bronislav Sidik, Chaya Levi, Joseph Kampeas
Published: 2026-01-29
Categories: cs.DC, cs.AI
arXiv: https://arxiv.org/abs/2601.21758v1

Serving Large Language Models (LLMs) under mixed workloads--short, latency-sensitive interactive queries alongside long, throughput-oriented batch requests--poses a fundamental scheduling challenge. Standard First-Come, First-Served (FCFS) policies suffer from severe head-of-line blocking, leading to high tail latency and underutilized hardware. We introduce EWSJF (Effective Workload-based Shortest Job First), an adaptive request-level scheduler that learns workload structure in real time to jointly improve fairness and throughput. EWSJF operates upstream of execution-level schedulers and integrates four components: (1) Refine-and-Prune, an unsupervised partitioning algorithm that discovers performance-homogeneous request groups; (2) Dynamic Queue Routing for assigning requests to these groups; (3) Density-Weighted Scoring, a context-aware prioritization function balancing urgency and fairness; and (4) Bayesian Meta-Optimization, which continuously tunes scoring and partitioning parameters based on live performance feedback. Implemented in vLLM, EWSJF improves end-to-end throughput by over 30% and reduces average Time-To-First-Token for short requests by up to 4x compared to FCFS. These results demonstrate that adaptive, learning-based request scheduling is a critical missing layer for efficient and responsive LLM serving. Implementation available at https://anonymous.4open.science/r/vllm_0110-32D8.

---

### Language-based Trial and Error Falls Behind in the Era of Experience

Authors: Haoyu Wang, Guozheng Ma, Shugang Cui, Yilun Kong, Haotian Luo, Li Shen, Mengya Gao, Yichao Wu, Xiaogang Wang, Dacheng Tao
Published: 2026-01-29
Categories: cs.AI
arXiv: https://arxiv.org/abs/2601.21754v3

While Large Language Models (LLMs) excel in language-based agentic tasks, their applicability to unseen, nonlinguistic environments (e.g., symbolic or spatial tasks) remains limited. Previous work attributes this performance gap to the mismatch between the pretraining distribution and the testing distribution. In this work, we demonstrate the primary bottleneck is the prohibitive cost of exploration: mastering these tasks requires extensive trial-and-error, which is computationally unsustainable for parameter-heavy LLMs operating in a high dimensional semantic space. To address this, we propose SCOUT (Sub-Scale Collaboration On Unseen Tasks), a novel framework that decouples exploration from exploitation. We employ lightweight "scouts" (e.g., small MLPs) to probe environmental dynamics at a speed and scale far exceeding LLMs. The collected trajectories are utilized to bootstrap the LLM via Supervised Fine-Tuning (SFT), followed by multi-turn Reinforcement Learning (RL) to activate its latent world knowledge. Empirically, SCOUT enables a Qwen2.5-3B-Instruct model to achieve an average score of 0.86, significantly outperforming proprietary models, including Gemini-2.5-Pro (0.60), while saving about 60% GPU hours consumption.

---

### Dynamic Topology Awareness: Breaking the Granularity Rigidity in Vision-Language Navigation

Authors: Jiankun Peng, Jianyuan Guo, Ying Xu, Yue Liu, Jiashuang Yan, Xuanwei Ye, Houhua Li, Xiaoming Wang
Published: 2026-01-29
Categories: cs.CV
arXiv: https://arxiv.org/abs/2601.21751v1

Vision-Language Navigation in Continuous Environments (VLN-CE) presents a core challenge: grounding high-level linguistic instructions into precise, safe, and long-horizon spatial actions. Explicit topological maps have proven to be a vital solution for providing robust spatial memory in such tasks. However, existing topological planning methods suffer from a "Granularity Rigidity" problem. Specifically, these methods typically rely on fixed geometric thresholds to sample nodes, which fails to adapt to varying environmental complexities. This rigidity leads to a critical mismatch: the model tends to over-sample in simple areas, causing computational redundancy, while under-sampling in high-uncertainty regions, increasing collision risks and compromising precision. To address this, we propose DGNav, a framework for Dynamic Topological Navigation, introducing a context-aware mechanism to modulate map density and connectivity on-the-fly. Our approach comprises two core innovations: (1) A Scene-Aware Adaptive Strategy that dynamically modulates graph construction thresholds based on the dispersion of predicted waypoints, enabling "densification on demand" in challenging environments; (2) A Dynamic Graph Transformer that reconstructs graph connectivity by fusing visual, linguistic, and geometric cues into dynamic edge weights, enabling the agent to filter out topological noise and enhancing instruction adherence. Extensive experiments on the R2R-CE and RxR-CE benchmarks demonstrate DGNav exhibits superior navigation performance and strong generalization capabilities. Furthermore, ablation studies confirm that our framework achieves an optimal trade-off between navigation efficiency and safe exploration. The code is available at https://github.com/shannanshouyin/DGNav.

---

### FISMO: Fisher-Structured Momentum-Orthogonalized Optimizer

Authors: Chenrui Xu, Wenjing Yan, Ying-Jun Angela Zhang
Published: 2026-01-29
Categories: cs.LG
arXiv: https://arxiv.org/abs/2601.21750v1

Training large-scale neural networks requires solving nonconvex optimization where the choice of optimizer fundamentally determines both convergence behavior and computational efficiency. While adaptive methods like Adam have long dominated practice, the recently proposed Muon optimizer achieves superior performance through orthogonalized momentum updates that enforce isotropic geometry with uniform singular values. However, this strict isotropy discards potentially valuable curvature information encoded in gradient spectra, motivating optimization methods that balance geometric structure with adaptivity. We introduce FISMO (Fisher-Structured Momentum-Orthogonalized) optimizer, which generalizes isotropic updates to incorporate anisotropic curvature information through Fisher information geometry. By reformulating the optimizer update as a trust-region problem constrained by a Kronecker-factored Fisher metric, FISMO achieves structured preconditioning that adapts to local loss landscape geometry while maintaining computational tractability. We establish convergence guarantees for FISMO in stochastic nonconvex settings, proving an $\mathcal{O}(1/\sqrt{T})$ rate for the expected squared gradient norm with explicit characterization of variance reduction through mini-batching. Empirical evaluation on image classification and language modeling benchmarks demonstrates that FISMO achieves superior training efficiency and final performance compared to established baselines.

---

### Temporal Sepsis Modeling: a Relational and Explainable-by-Design Framework

Authors: Vincent Lemaire, Nédra Meloulli, Pierre Jaquet
Published: 2026-01-29
Categories: cs.LG, cs.AI
arXiv: https://arxiv.org/abs/2601.21747v4

Sepsis remains one of the most complex and heterogeneous syndromes in intensive care. While deep learning models achieve competitive performance in early sepsis prediction, their decision processes often remain difficult to interpret clinically, and explainability is typically added only through post-hoc methods. We propose an explainable-by-design framework based on a relational approach: temporal EHR data are represented in a relational schema, flattened via MDL-based propositionalisation into compact human-readable features, and classified using a selective Fractional Naive Bayes classifier. Evaluated on MIMIC-III (3,940 patients, 10-fold cross-validation), our approach achieves AUC = 0.983 - competitive with XGBoost (0.985) and CatBoost (0.985), and superior to LSTM (0.945) - with only 98 selected variables and a 1 MB model footprint. Unlike post-hoc methods, interpretability is native and fourfold: univariate, global, local, and counterfactual.

---

### Temporal Guidance for Large Language Models

Authors: Hong-Kai Zheng, Piji Li
Published: 2026-01-29
Categories: cs.CL
arXiv: https://arxiv.org/abs/2601.21744v1

Contrastive Decoding (CD) enhances the generation quality of large language models (LLMs) but incurs significant additional computational overhead due to the need for an auxiliary model. Existing internal self-contrastive decoding methods, such as Decoding by Contrasting Layers (DoLa), focus on discrepancies across different layers, which are notably unstable on small-scale models. In this work, based on the observation that LLMs exhibit local preferences, we propose a novel contrastive guidance strategy along the temporal dimension, namely Temporal Guidance (TeGu). Our method ingeniously leverages Multi-Token Prediction (MTP) to construct weaker amateur predictions for model self-contrast. To standardize the implementation of this mechanism, we further introduce a lightweight Conditional MTP Projector (cMTPP), which avoids maintaining multiple independent networks as required by other MTP modules. Across various model series and benchmarks, TeGu achieves significant performance improvements while maintaining low additional memory consumption and computational overhead.

---

### Epistemic Context Learning: Building Trust the Right Way in LLM-Based Multi-Agent Systems

Authors: Ruiwen Zhou, Maojia Song, Xiaobao Wu, Sitao Cheng, Xunjian Yin, Yuxi Xie, Zhuoqun Hao, Wenyue Hua, Liangming Pan, Soujanya Poria, Min-Yen Kan
Published: 2026-01-29
Categories: cs.AI, cs.CL, cs.MA
arXiv: https://arxiv.org/abs/2601.21742v1

Individual agents in multi-agent (MA) systems often lack robustness, tending to blindly conform to misleading peers. We show this weakness stems from both sycophancy and inadequate ability to evaluate peer reliability. To address this, we first formalize the learning problem of history-aware reference, introducing the historical interactions of peers as additional input, so that agents can estimate peer reliability and learn from trustworthy peers when uncertain. This shifts the task from evaluating peer reasoning quality to estimating peer reliability based on interaction history. We then develop Epistemic Context Learning (ECL): a reasoning framework that conditions predictions on explicitly-built peer profiles from history. We further optimize ECL by reinforcement learning using auxiliary rewards. Our experiments reveal that our ECL enables small models like Qwen 3-4B to outperform a history-agnostic baseline 8x its size (Qwen 3-30B) by accurately identifying reliable peers. ECL also boosts frontier models to near-perfect (100%) performance. We show that ECL generalizes well to various MA configurations and we find that trust is modeled well by LLMs, revealing a strong correlation in trust modeling accuracy and final answer quality.

---

### Why $β_1 = β_2$ Is Dynamically Special in Adam

Authors: Alberto Fernández-Hernández, Cristian Pérez-Corral, Jose I. Mestre, Manuel F. Dolz, Enrique S. Quintana-Ortí
Published: 2026-01-29
Categories: cs.LG, cs.AI, stat.ML
arXiv: https://arxiv.org/abs/2601.21739v3

Adam has been at the core of large-scale training for almost a decade, yet the role of its two momentum parameters remains poorly understood. Recent work shows that tying $β_{1}=β_{2}$ can preserve Adam's strong performance despite collapsing two memory scales into one, raising a basic question: what becomes dynamically special when the memories are tied? We identify a concrete mechanism. In the continuous-time limit, each normalized-update coordinate decomposes into a sign component, an explicit magnitude-lag term proportional to the difference between the two memory times, and additional transition, curvature, and nonlinear ratio terms. This lag channel vanishes exactly when $β_{1}=β_{2}$, making the diagonal the unique regime in which this mismatch-induced response is structurally absent. A full-history discrete decomposition on real training gradients recovers this change in composition: tied updates are sign-dominated, whereas the lag term becomes substantial off the diagonal and leaves a comparatively small residual. Across six vision and language tasks, tied configurations also typically exhibit smoother update-norm trajectories. Overall, our results identify memory-scale mismatch as a concrete source of magnitude sensitivity in Adam and provide a mechanistic account of why tied momentum is dynamically distinctive.

---

### From Global to Granular: Revealing IQA Model Performance via Correlation Surface

Authors: Baoliang Chen, Danni Huang, Hanwei Zhu, Lingyu Zhu, Wei Zhou, Shiqi Wang, Yuming Fang, Weisi Lin
Published: 2026-01-29
Categories: cs.CV, cs.AI
arXiv: https://arxiv.org/abs/2601.21738v2

Evaluation of Image Quality Assessment (IQA) models has long been dominated by global correlation metrics, such as Pearson Linear Correlation Coefficient (PLCC) and Spearman Rank-Order Correlation Coefficient (SRCC). While widely adopted, these metrics reduce performance to a single scalar, failing to capture how ranking consistency varies across the local quality spectrum. For example, two IQA models may achieve identical SRCC values, yet one ranks high-quality images (related to high Mean Opinion Score, MOS) more reliably, while the other better discriminates image pairs with small quality/MOS differences (related to $|Δ$MOS$|$). Such complementary behaviors are invisible under global metrics. Moreover, SRCC and PLCC are sensitive to test-sample quality distributions, yielding unstable comparisons across test sets. To address these limitations, we propose \textbf{Granularity-Modulated Correlation (GMC)}, which provides a structured, fine-grained analysis of IQA performance. GMC includes: (1) a \textbf{Granularity Modulator} that applies Gaussian-weighted correlations conditioned on absolute MOS values and pairwise MOS differences ($|Δ$MOS$|$) to examine local performance variations, and (2) a \textbf{Distribution Regulator} that regularizes correlations to mitigate biases from non-uniform quality distributions. The resulting \textbf{correlation surface} maps correlation values as a joint function of MOS and $|Δ$MOS$|$, providing a 3D representation of IQA performance. Experiments on standard benchmarks show that GMC reveals performance characteristics invisible to scalar metrics, offering a more informative and reliable paradigm for analyzing, comparing, and deploying IQA models. Codes are available at https://github.com/Dniaaa/GMC.

---

### Mixed-Precision Training and Compilation for RRAM-based Computing-in-Memory Accelerators

Authors: Rebecca Pelke, Joel Klein, Jose Cubero-Cascante, Nils Bosbach, Jan Moritz Joseph, Rainer Leupers
Published: 2026-01-29
Categories: cs.LG, cs.ET
arXiv: https://arxiv.org/abs/2601.21737v3

Computing-in-Memory (CIM) accelerators are a promising solution for accelerating Machine Learning (ML) workloads, as they perform Matrix-Vector Multiplications (MVMs) on crossbar arrays directly in memory. Although the bit widths of the crossbar inputs and cells are very limited, most CIM compilers do not support quantization below 8 bit. As a result, a single MVM requires many compute cycles, and weights cannot be efficiently stored in a single crossbar cell. To address this problem, we propose a mixed-precision training and compilation framework for CIM architectures. The biggest challenge is the massive search space, that makes it difficult to find good quantization parameters. This is why we introduce a reinforcement learning-based strategy to find suitable quantization configurations that balance latency and accuracy. In the best case, our approach achieves up to a 2.48x speedup over existing state-of-the-art solutions, with an accuracy loss of only 0.086 %.

---

### CE-GOCD: Central Entity-Guided Graph Optimization for Community Detection to Augment LLM Scientific Question Answering

Authors: Jiayin Lan, Jiaqi Li, Baoxin Wang, Ming Liu, Dayong Wu, Shijin Wang, Bing Qin, Guoping Hu
Published: 2026-01-29
Categories: cs.CL
arXiv: https://arxiv.org/abs/2601.21733v1

Large Language Models (LLMs) are increasingly used for question answering over scientific research papers. Existing retrieval augmentation methods often rely on isolated text chunks or concepts, but overlook deeper semantic connections between papers. This impairs the LLM's comprehension of scientific literature, hindering the comprehensiveness and specificity of its responses. To address this, we propose Central Entity-Guided Graph Optimization for Community Detection (CE-GOCD), a method that augments LLMs' scientific question answering by explicitly modeling and leveraging semantic substructures within academic knowledge graphs. Our approach operates by: (1) leveraging paper titles as central entities for targeted subgraph retrieval, (2) enhancing implicit semantic discovery via subgraph pruning and completion, and (3) applying community detection to distill coherent paper groups with shared themes. We evaluated the proposed method on three NLP literature-based question-answering datasets, and the results demonstrate its superiority over other retrieval-augmented baseline approaches, confirming the effectiveness of our framework.

---

### Mechanistic Evidence for Spectral Structures in Prior-Data Fitted Networks

Authors: Kaustubh Sharma, Srijan Tiwari, Ojasva Nema, Parikshit Pareek
Published: 2026-01-29
Categories: cs.LG
arXiv: https://arxiv.org/abs/2601.21731v2

Prior-Data Fitted Networks (PFNs) enable amortized Bayesian inference in a single forward pass, yet their internal representations remain opaque. It is unknown whether PFNs encode identifiable Bayesian structure or merely memorize input-output mappings. We provide mechanistic evidence that PFNs learn structured spectral representations and that these can be extracted as explicit kernels. First, probing experiments across three architectures, including the publicly released TabPFN, show that spectral information is linearly decodable from the latent attention score and organized along a dominant principal axis. Activation patching and targeted subspace interventions establish that this information is causally used for prediction and concentrated in a low-dimensional subspace, with spectral directions an order of magnitude more effective than random ones. Crucially, these properties hold on TabPFN with both synthetic out-of-distribution inputs and real-world time series (Airline Passengers, Milk Production), indicating they are emergent features of PFN-style amortization over continuous regression tasks rather than artifacts of training prior. Second, we introduce a Filter Bank Decoder that maps frozen PFN latents to explicit spectral densities, reconstructing stationary kernels via Bochner's theorem. The resulting kernels support GP regression competitive with iterative baselines while requiring only a single forward pass, demonstrating that PFN priors are not merely implicit but are explicitly recoverable as portable Bayesian objects.

---

### DropoutTS: Sample-Adaptive Dropout for Robust Time Series Forecasting

Authors: Siru Zhong, Yiqiu Liu, Zhiqing Cui, Zezhi Shao, Fei Wang, Qingsong Wen, Yuxuan Liang
Published: 2026-01-29
Categories: cs.AI
arXiv: https://arxiv.org/abs/2601.21726v2

Deep time series models are vulnerable to noisy data ubiquitous in real-world applications. Existing robustness strategies either prune data or rely on costly prior quantification, failing to balance effectiveness and efficiency. In this paper, we introduce DropoutTS, a model-agnostic plugin that shifts the paradigm from "what" to learn to "how much" to learn. DropoutTS employs a Sample-Adaptive Dropout mechanism: leveraging spectral sparsity to efficiently quantify instance-level noise via reconstruction residuals, it dynamically calibrates model learning capacity by mapping noise to adaptive dropout rates - selectively suppressing spurious fluctuations while preserving fine-grained fidelity. Extensive experiments across diverse noise regimes and open benchmarks show DropoutTS consistently boosts superior backbones' performance, delivering advanced robustness with negligible parameter overhead and no architectural modifications. Our code is available at https://github.com/CityMind-Lab/DropoutTS.

---

### Procedural Pretraining: Warming Up Language Models with Abstract Data

Authors: Liangze Jiang, Zachary Shinnick, Anton van den Hengel, Hemanth Saratchandran, Damien Teney
Published: 2026-01-29
Categories: cs.CL, cs.LG
arXiv: https://arxiv.org/abs/2601.21725v2

Pretraining language models directly on web-scale corpora is the de facto paradigm. We study an alternative where the model is initially exposed to abstract structured data to ease the subsequent acquisition of rich semantic knowledge, much like humans learning simple logic and mathematics before higher reasoning. We focus on procedural data, generated by formal languages and other simple algorithms, as such abstract data. We first diagnose the algorithmic skills that different forms of procedural data can improve, often significantly. For example, the accuracy of context recall (Needle-in-a-haystack) jumps from 10 to 98% when a model is pretrained on Dyck sequences (balanced brackets). Second, we study how these gains are reflected in pretraining larger models (up to 1.3B). We find that front-loading as little as 0.1 to 0.3% procedural data significantly outperforms standard pretraining on natural language, code, and informal mathematics (C4, CodeParrot, and DeepMind-Math datasets). Notably, this also enables the models to reach the same loss value with only 55/67/86% of the original data and thus a comparable reduction in FLOPs. Third, we explore the mechanisms behind the benefits and find that procedural pretraining instills non-trivial structure in both attention and MLP layers. The former is particularly important for structured domains (e.g. code), and the latter for language. Finally, we lay a path for combining multiple forms of procedural data. Our results show that procedural pretraining is a simple, lightweight means of improving performance and accelerating language model pretraining, ultimately suggesting the promise of disentangling knowledge acquisition from reasoning in LLMs.

---

### Enhancing Language Models for Robust Greenwashing Detection

Authors: Neil Heinrich Braun, Keane Ong, Rui Mao, Erik Cambria, Gianmarco Mengaldo
Published: 2026-01-29
Categories: cs.CL, cs.AI
arXiv: https://arxiv.org/abs/2601.21722v1

Sustainability reports are critical for ESG assessment, yet greenwashing and vague claims often undermine their reliability. Existing NLP models lack robustness to these practices, typically relying on surface-level patterns that generalize poorly. We propose a parameter-efficient framework that structures LLM latent spaces by combining contrastive learning with an ordinal ranking objective to capture graded distinctions between concrete actions and ambiguous claims. Our approach incorporates gated feature modulation to filter disclosure noise and utilizes MetaGradNorm to stabilize multi-objective optimization. Experiments in cross-category settings demonstrate superior robustness over standard baselines while revealing a trade-off between representational rigidity and generalization.

---

### LoRA and Privacy: When Random Projections Help (and When They Don't)

Authors: Yaxi Hu, Johanna Düngler, Bernhard Schölkopf, Amartya Sanyal
Published: 2026-01-29
Categories: cs.LG, stat.ML
arXiv: https://arxiv.org/abs/2601.21719v1

We introduce the (Wishart) projection mechanism, a randomized map of the form $S \mapsto M f(S)$ with $M \sim W_d(1/r I_d, r)$ and study its differential privacy properties. For vector-valued queries $f$, we prove non-asymptotic DP guarantees without any additive noise, showing that Wishart randomness alone can suffice. For matrix-valued queries, however, we establish a sharp negative result: in the noise-free setting, the mechanism is not DP, and we demonstrate its vulnerability by implementing a near perfect membership inference attack (AUC $> 0.99$). We then analyze a noisy variant and prove privacy amplification due to randomness and low rank projection, in both large- and small-rank regimes, yielding stronger privacy guarantees than additive noise alone. Finally, we show that LoRA-style updates are an instance of the matrix-valued mechanism, implying that LoRA is not inherently private despite its built-in randomness, but that low-rank fine-tuning can be more private than full fine-tuning at the same noise level. Preliminary experiments suggest that tighter accounting enables lower noise and improved accuracy in practice.

---

### When Does Predictive Inverse Dynamics Outperform Behavior Cloning?

Authors: Lukas Schäfer, Pallavi Choudhury, Abdelhak Lemkhenter, Chris Lovett, Somjit Nath, Luis França, Matheus Ribeiro Furtado de Mendonça, Alex Lamb, Riashat Islam, Siddhartha Sen, John Langford, Katja Hofmann et al.
Published: 2026-01-29
Categories: cs.LG, cs.AI
arXiv: https://arxiv.org/abs/2601.21718v3

Behavior cloning (BC) is a practical offline imitation learning method, but it often fails when expert demonstrations are limited. Recent works have introduced a class of architectures named predictive inverse dynamics models (PIDMs) that combine a future-state predictor with an inverse dynamics model. While PIDMs often outperform BC, the reasons behind their benefits remain unclear. In this paper, we provide a theoretical explanation: PIDMs introduce a tradeoff. Conditioning the IDM on the predicted future state can significantly reduce variance, but the prediction itself introduces additional bias and variance. We establish conditions for PIDMs to achieve higher sample efficiency and lower prediction error than BC, with the gap widening when additional data sources are available. We validate the theoretical insights empirically in 2D navigation tasks, where BC requires up to five times (three times on average) more demonstrations than PIDM to reach comparable performance. Results are also illustrated in a complex 3D environment in a modern video game with high-dimensional visual inputs and stochastic transitions, where BC requires over 66\% more samples than PIDM.

---

### On sample complexity for covariance estimation via the unadjusted Langevin algorithm

Authors: Shogo Nakakita
Published: 2026-01-29
Categories: math.PR, stat.ML
arXiv: https://arxiv.org/abs/2601.21717v2

We establish sample complexity guarantees for estimating the covariance matrix of a strongly log-concave smooth distribution using the unadjusted Langevin algorithm (ULA). We quantitatively compare our complexity estimates on single-chain ULA with embarrassingly parallel ULA and derive that the sample complexity of the single-chain approach is smaller than that of embarrassingly parallel ULA by a logarithmic factor in the dimension and the reciprocal of the prescribed precision, with the difference arising from effective bias reduction through burn-in. The key technical contribution is a concentration bound for the sample covariance matrix around its expectation, derived via a log-Sobolev inequality for the joint distribution of ULA iterates.

---

### DreamActor-M2: Universal Character Image Animation via Spatiotemporal In-Context Learning

Authors: Mingshuang Luo, Shuang Liang, Zhengkun Rong, Yuxuan Luo, Tianshu Hu, Ruibing Hou, Hong Chang, Yong Li, Yuan Zhang, Mingyuan Gao
Published: 2026-01-29
Categories: cs.CV, cs.AI
arXiv: https://arxiv.org/abs/2601.21716v1

Character image animation aims to synthesize high-fidelity videos by transferring motion from a driving sequence to a static reference image. Despite recent advancements, existing methods suffer from two fundamental challenges: (1) suboptimal motion injection strategies that lead to a trade-off between identity preservation and motion consistency, manifesting as a "see-saw", and (2) an over-reliance on explicit pose priors (e.g., skeletons), which inadequately capture intricate dynamics and hinder generalization to arbitrary, non-humanoid characters. To address these challenges, we present DreamActor-M2, a universal animation framework that reimagines motion conditioning as an in-context learning problem. Our approach follows a two-stage paradigm. First, we bridge the input modality gap by fusing reference appearance and motion cues into a unified latent space, enabling the model to jointly reason about spatial identity and temporal dynamics by leveraging the generative prior of foundational models. Second, we introduce a self-bootstrapped data synthesis pipeline that curates pseudo cross-identity training pairs, facilitating a seamless transition from pose-dependent control to direct, end-to-end RGB-driven animation. This strategy significantly enhances generalization across diverse characters and motion scenarios. To facilitate comprehensive evaluation, we further introduce AW Bench, a versatile benchmark encompassing a wide spectrum of characters types and motion scenarios. Extensive experiments demonstrate that DreamActor-M2 achieves state-of-the-art performance, delivering superior visual fidelity and robust cross-domain generalization. Project Page: https://grisoon.github.io/DreamActor-M2/

---

### E-mem: Multi-agent based Episodic Context Reconstruction for LLM Agent Memory

Authors: Kaixiang Wang, Yidan Lin, Jiong Lou, Zhaojiacheng Zhou, Bunyod Suvonov, Jie Li
Published: 2026-01-29
Categories: cs.AI
arXiv: https://arxiv.org/abs/2601.21714v5

The evolution of Large Language Model (LLM) agents towards System~2 reasoning, characterized by deliberative, high-precision problem-solving, requires maintaining rigorous logical integrity over extended horizons. However, prevalent memory preprocessing paradigms suffer from destructive de-contextualization. By compressing complex sequential dependencies into pre-defined structures (e.g., embeddings or graphs), these methods sever the contextual integrity essential for deep reasoning. To address this, we propose E-mem, a framework shifting from Memory Preprocessing to Episodic Context Reconstruction. Inspired by biological engrams, E-mem employs a heterogeneous hierarchical architecture where multiple assistant agents maintain uncompressed memory contexts, while a central master agent orchestrates global planning. Unlike passive retrieval, our mechanism empowers assistants to locally reason within activated segments, extracting context-aware evidence before aggregation. Evaluations on the LoCoMo benchmark demonstrate that E-mem achieves over 54\% F1, surpassing the state-of-the-art GAM by 7.75\%, while reducing token cost by over 70\%.

---

### Disentangling perception and reasoning for improving data efficiency in learning cloth manipulation without demonstrations

Authors: Donatien Delehelle, Fei Chen, Darwin Caldwell
Published: 2026-01-29
Categories: cs.RO, cs.AI
arXiv: https://arxiv.org/abs/2601.21713v1

Cloth manipulation is a ubiquitous task in everyday life, but it remains an open challenge for robotics. The difficulties in developing cloth manipulation policies are attributed to the high-dimensional state space, complex dynamics, and high propensity to self-occlusion exhibited by fabrics. As analytical methods have not been able to provide robust and general manipulation policies, reinforcement learning (RL) is considered a promising approach to these problems. However, to address the large state space and complex dynamics, data-based methods usually rely on large models and long training times. The resulting computational cost significantly hampers the development and adoption of these methods. Additionally, due to the challenge of robust state estimation, garment manipulation policies often adopt an end-to-end learning approach with workspace images as input. While this approach enables a conceptually straightforward sim-to-real transfer via real-world fine-tuning, it also incurs a significant computational cost by training agents on a highly lossy representation of the environment state. This paper questions this common design choice by exploring an efficient and modular approach to RL for cloth manipulation. We show that, through careful design choices, model size and training time can be significantly reduced when learning in simulation. Furthermore, we demonstrate how the resulting simulation-trained model can be transferred to the real world. We evaluate our approach on the SoftGym benchmark and achieve significant performance improvements over available baselines on our task, while using a substantially smaller model.

---

### TACLer: Tailored Curriculum Reinforcement Learning for Efficient Reasoning

Authors: Huiyuan Lai, Malvina Nissim
Published: 2026-01-29
Categories: cs.CL, cs.AI
arXiv: https://arxiv.org/abs/2601.21711v1

Large Language Models (LLMs) have shown remarkable performance on complex reasoning tasks, especially when equipped with long chain-of-thought (CoT) reasoning. However, eliciting long CoT typically requires large-scale reinforcement learning (RL) training, while often leading to overthinking with redundant intermediate steps. To improve learning and reasoning efficiency, while preserving or even enhancing performance, we propose TACLer, a model-tailored curriculum reinforcement learning framework that gradually increases the complexity of the data based on the model's proficiency in multi-stage RL training. TACLer features two core components: (i) tailored curriculum learning that determines what knowledge the model lacks and needs to learn in progressive stages; (ii) a hybrid Thinking/NoThinking reasoning paradigm that balances accuracy and efficiency by enabling or disabling the Thinking mode. Our experiments show that TACLer yields a twofold advantage in learning and reasoning: (i) it reduces computational cost, cutting training compute by over 50% compared to long thinking models and reducing inference token usage by over 42% relative to the base model; and (ii) it improves accuracy by over 9% on the base model, consistently outperforming state-of-the-art Nothinking and Thinking baselines across four math datasets with complex problems.

---

### Why Attention Patterns Exist: A Unifying Temporal Perspective Analysis

Authors: Qingyue Yang, Jie Wang, Xing Li, Yinqi Bai, Xialiang Tong, Huiling Zhen, Jianye Hao, Mingxuan Yuan, Bin Li
Published: 2026-01-29
Categories: cs.CL
arXiv: https://arxiv.org/abs/2601.21709v1

Attention patterns play a crucial role in both training and inference of large language models (LLMs). Prior works have identified individual patterns such as retrieval heads, sink heads, and diagonal traces, yet these observations remain fragmented and lack a unifying explanation. To bridge this gap, we introduce \textbf{Temporal Attention Pattern Predictability Analysis (TAPPA), a unifying framework that explains diverse attention patterns by analyzing their underlying mathematical formulations} from a temporally continuous perspective. TAPPA both deepens the understanding of attention behavior and guides inference acceleration approaches. Specifically, TAPPA characterizes attention patterns as predictable patterns with clear regularities and unpredictable patterns that appear effectively random. Our analysis further reveals that this distinction can be explained by the degree of query self-similarity along the temporal dimension. Focusing on the predictable patterns, we further provide a detailed mathematical analysis of three representative cases through the joint effect of queries, keys, and Rotary Positional Embeddings (RoPE). We validate TAPPA by applying its insights to KV cache compression and LLM pruning tasks. Across these tasks, a simple metric motivated by TAPPA consistently improves performance over baseline methods. The code is available at https://github.com/MIRALab-USTC/LLM-TAPPA.

---

### FBS: Modeling Native Parallel Reading inside a Transformer

Authors: Tongxi Wang
Published: 2026-01-29
Categories: cs.AI, cs.CL
arXiv: https://arxiv.org/abs/2601.21708v2

Large language models (LLMs) excel across many tasks, yet inference is still dominated by strictly token-by-token autoregression. Existing acceleration methods largely patch this pipeline and miss core human-reading ingredients: content-adaptive foresight, chunk-structure-aware compute allocation, and train-test consistency for preview/skimming. We propose the Fovea-Block-Skip Transformer (FBS), which injects a causal, trainable loop into Transformers via Parafovea-Attention Window (PAW), Chunk-Head (CH), and Skip-Gate (SG). Across diverse benchmarks, FBS improves the quality-efficiency trade-off without increasing parameters, and ablations show the three modules are complementary.

---

### SmartMeterFM: Unifying Smart Meter Data Generative Tasks Using Flow Matching Models

Authors: Nan Lin, Yanbo Wang, Jacco Heres, Peter Palensky, Pedro P. Vergara
Published: 2026-01-29
Categories: cs.LG, eess.SY
arXiv: https://arxiv.org/abs/2601.21706v2

Smart meter data is the foundation for planning and operating the distribution network. Unfortunately, such data are not always available due to privacy regulations. Meanwhile, the collected data may be corrupted due to sensor or transmission failure, or it may not have sufficient resolution for downstream tasks. A wide range of generative tasks is formulated to address these issues, including synthetic data generation, missing data imputation, and super-resolution. Despite the success of machine learning models on these tasks, dedicated models need to be designed and trained for each task, leading to redundancy and inefficiency. In this paper, by recognizing the powerful modeling capability of flow matching models, we propose a new approach to unify diverse smart meter data generative tasks with a single model trained for conditional generation. The proposed flow matching models are trained to generate challenging, high-dimensional time series data, specifically monthly smart meter data at a 15 min resolution. By viewing different generative tasks as distinct forms of partial data observations and injecting them into the generation process, we unify tasks such as imputation and super-resolution with a single model, eliminating the need for re-training. The data generated by our model not only are consistent with the given observations but also remain realistic, showing better performance against interpolation and other machine learning based baselines dedicated to the tasks.

---

### Beyond Forgetting: Representation Misdirection Elicits Controllable Side Behaviors and Capabilities

Authors: Tien Dang, The-Hai Nguyen, Dinh Mai Phuong, Nguyen Minh Phuong, Anh Bui, Hoang Thanh-Tung, Le-Minh Nguyen, Naoya Inoue
Published: 2026-01-29
Categories: cs.LG, cs.CL
arXiv: https://arxiv.org/abs/2601.21702v4

We consider Representation Misdirection (RM), a class of large language model (LLM) unlearning methods that achieve forgetting by redirecting the latent representations of forget-samples toward a target vector. Despite being important, the roles of the target vector used in RM, however, remain underexplored. Here, we approach and revisit RM through the lens of the Linear Representation Hypothesis. Specifically, if one can identify a one-dimensional representation corresponding to a high-level concept, the Linear Representation Hypothesis enables linear operations on this concept vector within the forget-representation space. Under this view, we hypothesize that, beyond forgetting, machine unlearning via RM elicits controllable side effect behaviors and capabilities corresponding to the high-level concept. Our hypothesis is empirically validated across a wide range of concepts and tasks, including controlling unlearned models' truthfulness, sentiment, stereotypical bias, refusal, language, and in-context learning (ICL) tasks. Our findings reveal that this phenomenon could be either a hidden risk if misused or a mechanism that can be harnessed for developing unlearned models that require stronger capabilities and controllable behaviors.

---

### Toward Culturally Aligned LLMs through Ontology-Guided Multi-Agent Reasoning

Authors: Wonduk Seo, Wonseok Choi, Junseo Koh, Juhyeon Lee, Hyunjin An, Minhyeong Yu, Jian Park, Qingshan Zhou, Seunghyun Lee, Yi Bu
Published: 2026-01-29
Categories: cs.CL, cs.AI, cs.IR, cs.MA, cs.SI
arXiv: https://arxiv.org/abs/2601.21700v3

Large Language Models (LLMs) increasingly support culturally sensitive decision making, yet often exhibit misalignment due to skewed pretraining data and the absence of structured value representations. Existing methods can steer outputs, but often lack demographic grounding and treat values as independent, unstructured signals, reducing consistency and interpretability. We propose OG-MAR, an Ontology-Guided Multi-Agent Reasoning framework. OG-MAR summarizes respondent-specific values from the World Values Survey (WVS) and constructs a global cultural ontology by eliciting relations over a fixed taxonomy via competency questions. At inference time, it retrieves ontology-consistent relations and demographically similar profiles to instantiate multiple value-persona agents, whose outputs are synthesized by a judgment agent that enforces ontology consistency and demographic proximity. Experiments on regional social-survey benchmarks across four LLM backbones show that OG-MAR improves cultural alignment and robustness over competitive baselines, while producing more transparent reasoning traces.

---

### Can David Beat Goliath? On Multi-Hop Reasoning with Resource-Constrained Agents

Authors: Hojae Han, Heeyun Jung, Jongyoon Kim, Seung-won Hwang
Published: 2026-01-29
Categories: cs.CL, cs.IR
arXiv: https://arxiv.org/abs/2601.21699v4

Reinforcement learning (RL) trains small language model agents to answer multi-hop questions by retrieving evidence over multiple turns, but reported gains typically rely on thousands of on-policy rollouts per update. We study RL for such agents under the budget constraint of commodity GPUs, where each update samples only a few rollouts per question. Under this constraint, most sampled trajectories retrieve none of the required evidence, so the outcome reward gives the policy little to learn from and small agents settle for answering without retrieval, a failure we call \emph{retrieval collapse}. David-GRPO addresses this with two mechanisms: (1) \emph{Expert trajectory seeding} places a handful of off-policy expert trajectories into the GRPO groups of the early updates, and (2) \emph{evidence-guided continuation} rewards evidence coverage and resumes the most promising partial trajectory. The evidence for each training question is constructed from the corpus link graph, so no annotated evidence is required. On six multi-hop QA benchmarks, David-GRPO trained on four RTX 3090 GPUs with 144 rollouts per step brings Qwen2.5-1.5B to 22.6 average EM against 11.9 for the best baseline under the same budget, matches Tree-GRPO trained with 20 times more rollouts, and, unlike the baselines that stop after at most one search, learns to retrieve across turns. The implementation is available at: https://github.com/AsadalJung/David-GRPO

---

### Curriculum Learning for LLM Pretraining: An Analysis of Learning Dynamics

Authors: Mohamed Elgaar, Hadi Amiri
Published: 2026-01-29
Categories: cs.LG, cs.AI
arXiv: https://arxiv.org/abs/2601.21698v2

Curriculum learning changes the order of pretraining data, but it remains unclear how ordering changes the learning dynamics. We pretrain models from 14M to 1B parameters for 300B tokens under three linguistically motivated curricula--Age-of-Acquisition, word frequency, and Verb Variation (VV)--and compare each against Random ordering. We analyze latent training phases, gradient noise scale (GNS), and the singular-value structure of the output head. We find that training follows a shared sequence of latent phases, while curricula mainly change time spent in each phase. Random ordering yields higher GNS at 14M-70M and late singular-entropy spikes up to 160M, consistent with noisier gradients and output-head saturation. A reverse-order VV control shows that direction matters: descending order loses much of the accuracy advantage of the ascending curriculum. At larger scales, these stability differences are smaller. These results indicate that the curricula studied here are associated with more stable within-phase training in smaller models rather than with the creation of new phases.

---

### Independent Component Discovery in Temporal Count Data

Authors: Alexandre Chaussard, Anna Bonnet, Sylvain Le Corff
Published: 2026-01-29
Categories: stat.ME, stat.ML
arXiv: https://arxiv.org/abs/2601.21696v2

Advances in data collection are producing growing volumes of temporal count observations, making adapted modeling increasingly necessary. In this work, we introduce a generative framework for independent component analysis of temporal count data, combining regime-adaptive dynamics with Poisson log-normal emissions. The model identifies disentangled components with regime-dependent contributions, enabling representation learning and perturbations analysis. Notably, we establish the identifiability of the model, supporting principled interpretation. To learn the parameters, we propose an efficient amortized variational inference procedure. Experiments on simulated data evaluate recovery of the mixing function and latent sources across diverse settings, while real-world applications to gut microbiome and climate datasets reveal co-variation patterns and regime shifts consistent with domain-specific knowledge.

---

### ChartE$^{3}$: A Comprehensive Benchmark for End-to-End Chart Editing

Authors: Shuo Li, Jiajun Sun, Zhekai Wang, Xiaoran Fan, Hui Li, Dingwen Yang, Zhiheng Xi, Yijun Wang, Zifei Shan, Tao Gui, Qi Zhang, Xuanjing Huang
Published: 2026-01-29
Categories: cs.CV
arXiv: https://arxiv.org/abs/2601.21694v1

Charts are a fundamental visualization format for structured data analysis. Enabling end-to-end chart editing according to user intent is of great practical value, yet remains challenging due to the need for both fine-grained control and global structural consistency. Most existing approaches adopt pipeline-based designs, where natural language or code serves as an intermediate representation, limiting their ability to faithfully execute complex edits. We introduce ChartE$^{3}$, an End-to-End Chart Editing benchmark that directly evaluates models without relying on intermediate natural language programs or code-level supervision. ChartE$^{3}$ focuses on two complementary editing dimensions: local editing, which involves fine-grained appearance changes such as font or color adjustments, and global editing, which requires holistic, data-centric transformations including data filtering and trend line addition. ChartE$^{3}$ contains over 1,200 high-quality samples constructed via a well-designed data pipeline with human curation. Each sample is provided as a triplet of a chart image, its underlying code, and a multimodal editing instruction, enabling evaluation from both objective and subjective perspectives. Extensive benchmarking of state-of-the-art multimodal large language models reveals substantial performance gaps, particularly on global editing tasks, highlighting critical limitations in current end-to-end chart editing capabilities.

---

### TCAP: Tri-Component Attention Profiling for Unsupervised Backdoor Detection in MLLM Fine-Tuning

Authors: Mingzu Liu, Hao Fang, Runmin Cong
Published: 2026-01-29
Categories: cs.AI
arXiv: https://arxiv.org/abs/2601.21692v2

Fine-Tuning-as-a-Service (FTaaS) facilitates the customization of Multimodal Large Language Models (MLLMs) but introduces critical backdoor risks via poisoned data. Existing defenses either rely on supervised signals or fail to generalize across diverse trigger types and modalities. In this work, we uncover a universal backdoor fingerprint-attention allocation divergence-where poisoned samples disrupt the balanced attention distribution across three functional components: system instructions, vision inputs, and user textual queries, regardless of trigger morphology. Motivated by this insight, we propose Tri-Component Attention Profiling (TCAP), an unsupervised defense framework to filter backdoor samples. TCAP decomposes cross-modal attention maps into the three components, identifies trigger-responsive attention heads via Gaussian Mixture Model (GMM) statistical profiling, and isolates poisoned samples through EM-based vote aggregation. Extensive experiments across diverse MLLM architectures and attack methods demonstrate that TCAP achieves consistently strong performance, establishing it as a robust and practical backdoor defense in MLLMs.

---

### A Unified Generalization Framework for Model Merging: Trade-offs, Non-Linearity, and Scaling Laws

Authors: Qinglun Li, Anke Tang, Miao Zhang, Mengzhu Wang, Quanjun Yin, Li Shen
Published: 2026-01-29
Categories: cs.LG
arXiv: https://arxiv.org/abs/2601.21690v2

Model merging efficiently aggregates capabilities from multiple fine-tuned models into a single one, operating purely in parameter space without original data or expensive re-computation. Despite empirical successes, a unified theory for its effectiveness under heterogeneous finetuning hyperparameters (e.g., varying learning rates, batch sizes) remains missing. Existing federated learning theories focus purely on optimization, which fails to explain model merging and inherently leads to theoretical paradoxes. To address this challenge, we pioneer the integration of $L_2$-Stability theory into heterogeneous environments to rigorously decouple the excess risk of the merged model $\boldsymbol{x}_{avg}$ into optimization and generalization errors. This comprehensive analysis yields three main contributions: (i) We mathematically establish the fundamental \textit{Optimization-Generalization Trade-off}, explicitly resolving the paradox of why over-trained experts lead to catastrophic merging collapse. (ii) \textit{A unified theoretical framework} is provided to explain not only linear merging algorithms (e.g., TA, AdaMerging) but also state-of-the-art \textit{non-linear} merging algorithms (e.g., TIES, DARE), proving how sparsification operators strictly tighten the generalization bound by suppressing task heterogeneity. (iii) Rather than heuristic guidelines, we derive \textit{Quantitative Scaling Laws} that theoretically predict the precise impact of hyperparameter choices, enabling practitioners to strategically construct ``merge-friendly'' experts. Extensive experiments on the ResNet and ViT architectures across 20 visual classification tasks, involving thousands of finetuning models, robustly confirm that our theoretical scaling laws accurately predict the empirical generalization behaviors of $\boldsymbol{x}_{avg}$.

---