# Research Papers — 2026-02 · bundle page 112 of 210

> Full-text markdown of papers #5551–5600 of 10493 published in 2026-02, newest first. AI-readable mirrors of the full 10493-paper 2026-02 archive (50 papers per file).

Canonical page: https://aipod.dev/papers/archive/2026-02/page/112
[← Previous bundle](https://aipod.dev/papers/archive/2026-02/page/111.md)
[Next bundle →](https://aipod.dev/papers/archive/2026-02/page/113.md)

### Patch the Distribution Mismatch: RL Rewriting Agent for Stable Off-Policy SFT

Authors: Jiacheng Wang, Zhijie Liu, Ping Jian, Zirong Chen, Ke Ren Liao, Zhen Yang, Zhongbin Guo
Published: 2026-02-11
Categories: cs.LG, cs.CL
arXiv: https://arxiv.org/abs/2602.11220v2

Large language models are commonly adapted to downstream tasks through supervised fine-tuning (SFT), but substantial distribution mismatch between downstream supervision and a model's generation distribution can intensify catastrophic forgetting. Data rewriting offers a data-centric way to narrow this mismatch before SFT. Existing methods, however, typically sample rewrites from a prompt-induced conditional distribution, which need not align with the backbone's natural question-answering generation distribution, and fixed templates can reduce output diversity. We formulate data rewriting as a policy-learning problem and train a lightweight LoRA rewriting policy with reinforcement learning. The policy optimizes question-answering-style distributional alignment and semantic diversity under a hard task-consistency gate, producing verified supervision for downstream SFT. Across three instruction-tuned backbones, the resulting models attain downstream gains broadly comparable to standard SFT while reducing degradation on non-downstream benchmarks in every evaluated setting. Additional experiments on logical reasoning and medical question answering provide preliminary evidence that a rewriting policy can be reused across domains for the same backbone.

---

### Patch the Distribution Mismatch: RL Rewriting Agent for Stable Off-Policy SFT

Authors: Jiacheng Wang, Ping Jian, Zhen Yang, Zirong Chen, Keren Liao, Zhongbin Guo
Published: 2026-02-11
Categories: cs.LG, cs.CL
arXiv: https://arxiv.org/abs/2602.11220v1

Large language models (LLMs) have made rapid progress, yet adapting them to downstream scenarios still commonly relies on supervised fine-tuning (SFT). When downstream data exhibit a substantial distribution shift from the model's prior training distribution, SFT can induce catastrophic forgetting. To narrow this gap, data rewriting has been proposed as a data-centric approach that rewrites downstream training data prior to SFT. However, existing methods typically sample rewrites from a prompt-induced conditional distribution, so the resulting targets are not necessarily aligned with the model's natural QA-style generation distribution. Moreover, reliance on fixed templates can lead to diversity collapse. To address these issues, we cast data rewriting as a policy learning problem and learn a rewriting policy that better matches the backbone's QA-style generation distribution while preserving diversity. Since distributional alignment, diversity and task consistency are automatically evaluable but difficult to optimize end-to-end with differentiable objectives, we leverage reinforcement learning to optimize the rewrite distribution under reward feedback and propose an RL-based data-rewriting agent. The agent jointly optimizes QA-style distributional alignment and diversity under a hard task-consistency gate, thereby constructing a higher-quality rewritten dataset for downstream SFT. Extensive experiments show that our method achieves downstream gains comparable to standard SFT while reducing forgetting on non-downstream benchmarks by 12.34% on average. Our code is available at https://anonymous.4open.science/r/Patch-the-Prompt-Gap-4112 .

---

### Structurally Separated Uncertainty in Supervised Latent Variable Models

Authors: Tanmoy Mukherjee, Marius Kloft, Pierre Marquis, Zied Bouraoui
Published: 2026-02-11
Categories: cs.LG, cs.AI
arXiv: https://arxiv.org/abs/2602.11219v2

Predictive uncertainty is commonly decomposed into epistemic and aleatoric components, but standard decompositions often produce strongly correlated estimates because both quantities are derived from the same predictive distribution. We study an alternative design principle, \emph{structural separation}, which assigns epistemic and aleatoric uncertainty to disjoint parameter paths trained with distinct supervision targets: reducible prediction error for epistemic uncertainty and persistent label ambiguity for aleatoric uncertainty. We instantiate this principle in two supervised latent-variable models, a Credal Concept Bottleneck Model and a credal Self-Explaining Neural Network, and prove a gradient-isolation result showing that the two uncertainty heads are not coupled through shared training gradients under the proposed parameterization. Across five ambiguity-aware benchmarks, structural separation substantially reduces epistemic-aleatoric correlation while preserving predictive performance. Further analyses show that aleatoric estimates track annotator- or corpus-derived ambiguity, while epistemic estimates are more sensitive to prediction error and data availability. These results suggest that supervised latent-variable architectures provide a practical route toward uncertainty estimates that are not merely decorrelated, but operationally distinguishable.

---

### The Magic Correlations: Understanding Knowledge Transfer from Pretraining to Supervised Fine-Tuning

Authors: Simin Fan, Dimitris Paparas, Natasha Noy, Binbin Xiong, Noveen Sachdeva, Berivan Isik
Published: 2026-02-11
Categories: cs.LG
arXiv: https://arxiv.org/abs/2602.11217v1

Understanding how language model capabilities transfer from pretraining to supervised fine-tuning (SFT) is fundamental to efficient model development and data curation. In this work, we investigate four core questions: RQ1. To what extent do accuracy and confidence rankings established during pretraining persist after SFT? RQ2. Which benchmarks serve as robust cross-stage predictors and which are unreliable? RQ3. How do transfer dynamics shift with model scale? RQ4. How well does model confidence align with accuracy, as a measure of calibration quality? Does this alignment pattern transfer across training stages? We address these questions through a suite of correlation protocols applied to accuracy and confidence metrics across diverse data mixtures and model scales. Our experiments reveal that transfer reliability varies dramatically across capability categories, benchmarks, and scales -- with accuracy and confidence exhibiting distinct, sometimes opposing, scaling dynamics. These findings shed light on the complex interplay between pretraining decisions and downstream outcomes, providing actionable guidance for benchmark selection, data curation, and efficient model development.

---

### Protein Language Model Embeddings Improve Generalization of Implicit Transfer Operators

Authors: Panagiotis Antoniadis, Beatrice Pavesi, Simon Olsson, Ole Winther
Published: 2026-02-11
Categories: cs.LG, physics.bio-ph
arXiv: https://arxiv.org/abs/2602.11216v2

Molecular dynamics (MD) is a central computational tool in physics, chemistry, and biology, enabling quantitative prediction of experimental observables as expectations over high-dimensional molecular distributions such as Boltzmann distributions and transition densities. However, conventional MD is fundamentally limited by the high computational cost required to generate independent samples. Generative molecular dynamics (GenMD) has recently emerged as an alternative, learning surrogates of molecular distributions either from data or through interaction with energy models. While these methods enable efficient sampling, their transferability across molecular systems is often limited. In this work, we show that incorporating auxiliary sources of information can improve the data efficiency and generalization of transferable implicit transfer operators (TITO) for molecular dynamics. We find that coarse-grained TITO models are substantially more data-efficient than Boltzmann Emulators, and that incorporating protein language model (pLM) embeddings further improves out-of-distribution generalization. Our approach, PLaTITO, achieves state-of-the-art performance on equilibrium sampling benchmarks for out-of-distribution protein systems, including fast-folding proteins. We further study the impact of additional conditioning signals such as structural embeddings, temperature, and large-language-model-derived embeddings on model performance.

---

### Charting Empirical Laws for LLM Fine-Tuning in Scientific Multi-Discipline Learning

Authors: Lintao Wang, Zhuqiang Lu, Yilin Zhu, Kun Hu, Zhenfei Yin, Shixiang Tang, Zhiyong Wang, Wanli Ouyang, Xinzhu Ma
Published: 2026-02-11
Categories: cs.LG
arXiv: https://arxiv.org/abs/2602.11215v1

While large language models (LLMs) have achieved strong performance through fine-tuning within individual scientific domains, their learning dynamics in multi-disciplinary contexts remains poorly understood, despite the promise of improved generalization and broader applicability through cross-domain knowledge synergy. In this work, we present the first systematic study of multi-disciplinary LLM fine-tuning, constructing a five-discipline corpus and analyzing learning patterns of full fine-tuning, LoRA, LoRA-MoE, and LoRA compositions. Particularly, our study shows that multi-disciplinary learning is substantially more variable than single-discipline training and distills four consistent empirical laws: (1) Balance-then-Diversity: low-resource disciplines degrade performance unless mitigated via diversity-aware upsampling; (2) Merge-then-Align: restoring instruction-following ability is critical for cross-discipline synergy; (3) Optimize-then-Scale: parameter scaling offers limited gains without prior design optimization; and (4) Share-then-Specialize: asymmetric LoRA-MoE yields robust gains with minimal trainable parameters via shared low-rank projection. Together, these laws form a practical recipe for principled multi-discipline fine-tuning and provide actionable guidance for developing generalizable scientific LLMs.

---

### DD-MDN: Human Trajectory Forecasting with Diffusion-Based Dual Mixture Density Networks and Uncertainty Self-Calibration

Authors: Manuel Hetzel, Kerim Turacan, Hannes Reichert, Konrad Doll, Bernhard Sick
Published: 2026-02-11
Categories: cs.CV, cs.RO
arXiv: https://arxiv.org/abs/2602.11214v1

Human Trajectory Forecasting (HTF) predicts future human movements from past trajectories and environmental context, with applications in Autonomous Driving, Smart Surveillance, and Human-Robot Interaction. While prior work has focused on accuracy, social interaction modeling, and diversity, little attention has been paid to uncertainty modeling, calibration, and forecasts from short observation periods, which are crucial for downstream tasks such as path planning and collision avoidance. We propose DD-MDN, an end-to-end probabilistic HTF model that combines high positional accuracy, calibrated uncertainty, and robustness to short observations. Using a few-shot denoising diffusion backbone and a dual mixture density network, our method learns self-calibrated residence areas and probability-ranked anchor paths, from which diverse trajectory hypotheses are derived, without predefined anchors or endpoints. Experiments on the ETH/UCY, SDD, inD, and IMPTC datasets demonstrate state-of-the-art accuracy, robustness at short observation intervals, and reliable uncertainty modeling. The code is available at: https://github.com/kav-institute/ddmdn.

---

### Towards Compressive and Scalable Recurrent Memory

Authors: Yunchong Song, Jushi Kai, Liming Lu, Kaixi Qiu, Zhouhan Lin
Published: 2026-02-11
Categories: cs.LG
arXiv: https://arxiv.org/abs/2602.11212v1

Transformers face a quadratic bottleneck in attention when scaling to long contexts. Recent approaches introduce recurrent memory to extend context beyond the current window, yet these often face a fundamental trade-off between theoretical principles and practical scalability. To address this, we introduce Elastic Memory, a novel memory architecture grounded in the HiPPO framework for online function approximation. Elastic Memory treats historical sequence as samples from continuous signals, applying optimal online compression to encode them into a fixed-size memory state. For retrieval, we propose a flexible \textit{polynomial sampling} mechanism that reconstructs a history summary from this compressed state. Elastic Memory consistently outperformed baselines on long-context (32k+) datasets across three domains. With equal parameters, it beat Memorizing Transformer by 16x memory and outperformed Melodi at all memory sizes, even when Melodi had 30% more parameters. When scaling model size, Elastic Memory stayed ahead of all baselines and was significantly faster than Melodi at 4x size. Furthermore, its decoupled design allows for injecting inductive biases at test-time to boost performance.

---

### SWE-MiniSandbox: Container-Free Reinforcement Learning for Building Software Engineering Agents

Authors: Danlong Yuan, Wei Wu, Enhan Zhao, Zhengren Wang, Xueliang Zhao, Huishuai Zhang, Dongyan Zhao
Published: 2026-02-11
Categories: cs.SE, cs.AI, cs.LG
arXiv: https://arxiv.org/abs/2602.11210v5

Reinforcement learning (RL) has become a key paradigm for training software engineering (SWE) agents, but existing pipelines typically rely on per-task containers for isolation. At scale, pre-built container images incur substantial storage overhead, slow environment setup, and require container-management privileges. We propose SWE-MiniSandbox, a lightweight, container-free method that enables scalable RL training of SWE agents without sacrificing isolation. Instead of relying on per-instance containers, SWE-MiniSandbox executes each task in an isolated workspace backed by kernel-level mechanisms, substantially reducing system overhead. It leverages lightweight environment pre-caching techniques to eliminate the need for bulky container images. As a result, our approach lowers disk usage to approximately 5\% of that required by container-based pipelines and reduces environment preparation time to about 25\% of the container baseline. Empirical results demonstrate that SWE-MiniSandbox achieves evaluation performance comparable to standard container-based pipelines. By removing the dependency on heavy container infrastructure, SWE-MiniSandbox offers a practical and accessible foundation for scaling RL-based SWE agents, particularly in resource-constrained research environments.

---

### SurfPhase: 3D Interfacial Dynamics in Two-Phase Flows from Sparse Videos

Authors: Yue Gao, Hong-Xing Yu, Sanghyeon Chang, Qianxi Fu, Bo Zhu, Yoonjin Won, Juan Carlos Niebles, Jiajun Wu
Published: 2026-02-11
Categories: cs.CV
arXiv: https://arxiv.org/abs/2602.11154v1

Interfacial dynamics in two-phase flows govern momentum, heat, and mass transfer, yet remain difficult to measure experimentally. Classical techniques face intrinsic limitations near moving interfaces, while existing neural rendering methods target single-phase flows with diffuse boundaries and cannot handle sharp, deformable liquid-vapor interfaces. We propose SurfPhase, a novel model for reconstructing 3D interfacial dynamics from sparse camera views. Our approach integrates dynamic Gaussian surfels with a signed distance function formulation for geometric consistency, and leverages a video diffusion model to synthesize novel-view videos to refine reconstruction from sparse observations. We evaluate on a new dataset of high-speed pool boiling videos, demonstrating high-quality view synthesis and velocity estimation from only two camera views. Project website: https://yuegao.me/SurfPhase.

---

### Diffusion-Pretrained Dense and Contextual Embeddings

Authors: Sedigheh Eslami, Maksim Gaiduk, Markus Krimmel, Louis Milliken, Bo Wang, Denis Bykov
Published: 2026-02-11
Categories: cs.LG, cs.CL, cs.IR
arXiv: https://arxiv.org/abs/2602.11151v2

In this report, we introduce pplx-embed, a family of multilingual embedding models that employ multi-stage contrastive learning on a diffusion-pretrained language model backbone for web-scale retrieval. By leveraging bidirectional attention through diffusion-based pretraining, our models capture comprehensive bidirectional context within passages, enabling the use of mean pooling and a late chunking strategy to better preserve global context across long documents. We release two model types: pplx-embed-v1 for standard retrieval, and pplx-embed-context-v1 for contextualized embeddings that incorporate global document context into passage representations. pplx-embed-v1 achieves competitive performance on the MTEB(Multilingual, v2), MTEB(Code), MIRACL, BERGEN, and ToolRet retrieval benchmarks, while pplx-embed-context-v1 sets new records on the ConTEB benchmark. Beyond public benchmarks, pplx-embed-v1 demonstrates strong performance on our internal evaluation suite, focusing on real-world, large-scale search scenarios constructed from 1B production web pages. These results validate the models' effectiveness in production environments where retrieval quality and efficiency are critical at scale.

---

### YOR: Your Own Mobile Manipulator for Generalizable Robotics

Authors: Manan H Anjaria, Mehmet Enes Erciyes, Vedant Ghatnekar, Neha Navarkar, Haritheja Etukuru, Xiaole Jiang, Kanad Patel, Dhawal Kabra, Nicholas Wojno, Radhika Ajay Prayage, Soumith Chintala, Lerrel Pinto et al.
Published: 2026-02-11
Categories: cs.RO, cs.LG
arXiv: https://arxiv.org/abs/2602.11150v1

Recent advances in robot learning have generated significant interest in capable platforms that may eventually approach human-level competence. This interest, combined with the commoditization of actuators, has propelled growth in low-cost robotic platforms. However, the optimal form factor for mobile manipulation, especially on a budget, remains an open question. We introduce YOR, an open-source, low-cost mobile manipulator that integrates an omnidirectional base, a telescopic vertical lift, and two arms with grippers to achieve whole-body mobility and manipulation. Our design emphasizes modularity, ease of assembly using off-the-shelf components, and affordability, with a bill-of-materials cost under 10,000 USD. We demonstrate YOR's capability by completing tasks that require coordinated whole-body control, bimanual manipulation, and autonomous navigation. Overall, YOR offers competitive functionality for mobile manipulation research at a fraction of the cost of existing platforms. Project website: https://www.yourownrobot.ai/

---

### Data Repetition Beats Data Scaling in Long-CoT Supervised Fine-Tuning

Authors: Dawid J. Kopiczko, Sagar Vaze, Tijmen Blankevoort, Yuki M. Asano
Published: 2026-02-11
Categories: cs.CL
arXiv: https://arxiv.org/abs/2602.11149v2

Supervised fine-tuning (SFT) on chain-of-thought data is an essential post-training step for reasoning language models. Standard machine learning intuition suggests that training with more unique training samples yields better generalization. Counterintuitively, we show that SFT benefits from repetition: under a fixed update budget, training for more epochs on smaller datasets outperforms single-epoch training on larger datasets. On AIME'24/25 and GPQA benchmarks, Olmo3-7B trained for 128 epochs on 400 samples outperforms the equivalent 1 epoch on 51200 samples by 12-26 percentage points, with no additional catastrophic forgetting. We find that training token accuracy reliably signals when repetition has saturated; improvements from additional epochs plateau at full memorization, a pattern consistent across all settings. These findings provide a practical approach for reasoning SFT, where scaling epochs with token accuracy as a stopping criterion can replace expensive undirected data scaling. We pose the repetition advantage, where full memorization coincides with improved generalization, as a new open problem for the community in understanding the training dynamics of large language models.

---

### Data Repetition Beats Data Scaling in Long-CoT Supervised Fine-Tuning

Authors: Dawid J. Kopiczko, Sagar Vaze, Tijmen Blankevoort, Yuki M. Asano
Published: 2026-02-11
Categories: cs.CL
arXiv: https://arxiv.org/abs/2602.11149v1

Supervised fine-tuning (SFT) on chain-of-thought data is an essential post-training step for reasoning language models. Standard machine learning intuition suggests that training with more unique training samples yields better generalization. Counterintuitively, we show that SFT benefits from repetition: under a fixed update budget, training for more epochs on smaller datasets outperforms single-epoch training on larger datasets. On AIME'24/25 and GPQA benchmarks, Olmo3-7B trained for 128 epochs on 400 samples outperforms the equivalent 1 epoch on 51200 samples by 12-26 percentage points, with no additional catastrophic forgetting. We find that training token accuracy reliably signals when repetition has saturated; improvements from additional epochs plateau at full memorization, a pattern consistent across all settings. These findings provide a practical approach for reasoning SFT, where scaling epochs with token accuracy as a stopping criterion can replace expensive undirected data scaling. We pose the repetition advantage, where full memorization coincides with improved generalization, as a new open problem for the community in understanding the training dynamics of large language models.

---

### Beyond VLM-Based Rewards: Diffusion-Native Latent Reward Modeling

Authors: Gongye Liu, Bo Yang, Yida Zhi, Zhizhou Zhong, Lei Ke, Didan Deng, Han Gao, Yongxiang Huang, Kaihao Zhang, Hongbo Fu, Wenhan Luo
Published: 2026-02-11
Categories: cs.CV, cs.AI
arXiv: https://arxiv.org/abs/2602.11146v2

Preference optimization for diffusion and flow-matching models relies on reward functions that are both discriminatively robust and computationally efficient. Vision-Language Models (VLMs) have emerged as the primary reward provider, leveraging their rich multimodal priors to guide alignment. However, their computation and memory cost can be substantial, and optimizing a latent diffusion generator through a pixel-space reward introduces a domain mismatch that complicates alignment. In this paper, we propose DiNa-LRM, a diffusion-native latent reward model that formulates preference learning directly on noisy diffusion states. Our method introduces a noise-calibrated Thurstone likelihood with diffusion-noise-dependent uncertainty. DiNa-LRM leverages a pretrained latent diffusion backbone with a timestep-conditioned reward head, and supports inference-time noise ensembling, providing a diffusion-native mechanism for test-time scaling and robust rewarding. Across image alignment benchmarks, DiNa-LRM substantially outperforms existing diffusion-based reward baselines and achieves performance competitive with state-of-the-art VLMs at a fraction of the computational cost. In preference optimization, we demonstrate that DiNa-LRM improves preference optimization dynamics, enabling faster and more resource-efficient model alignment.

---

### SCRAPL: Scattering Transform with Random Paths for Machine Learning

Authors: Christopher Mitcheltree, Vincent Lostanlen, Emmanouil Benetos, Mathieu Lagrange
Published: 2026-02-11
Categories: cs.SD, cs.LG, eess.AS
arXiv: https://arxiv.org/abs/2602.11145v1

The Euclidean distance between wavelet scattering transform coefficients (known as paths) provides informative gradients for perceptual quality assessment of deep inverse problems in computer vision, speech, and audio processing. However, these transforms are computationally expensive when employed as differentiable loss functions for stochastic gradient descent due to their numerous paths, which significantly limits their use in neural network training. Against this problem, we propose "Scattering transform with Random Paths for machine Learning" (SCRAPL): a stochastic optimization scheme for efficient evaluation of multivariable scattering transforms. We implement SCRAPL for the joint time-frequency scattering transform (JTFS) which demodulates spectrotemporal patterns at multiple scales and rates, allowing a fine characterization of intermittent auditory textures. We apply SCRAPL to differentiable digital signal processing (DDSP), specifically, unsupervised sound matching of a granular synthesizer and the Roland TR-808 drum machine. We also propose an initialization heuristic based on importance sampling, which adapts SCRAPL to the perceptual content of the dataset, improving neural network convergence and evaluation performance. We make our code and audio samples available and provide SCRAPL as a Python package.

---

### GENIUS: Generative Fluid Intelligence Evaluation Suite

Authors: Ruichuan An, Sihan Yang, Ziyu Guo, Wei Dai, Zijun Shen, Haodong Li, Renrui Zhang, Xinyu Wei, Guopeng Li, Wenshan Wu, Wentao Zhang
Published: 2026-02-11
Categories: cs.LG, cs.AI, cs.CV
arXiv: https://arxiv.org/abs/2602.11144v1

Unified Multimodal Models (UMMs) have shown remarkable progress in visual generation. Yet, existing benchmarks predominantly assess $\textit{Crystallized Intelligence}$, which relies on recalling accumulated knowledge and learned schemas. This focus overlooks $\textit{Generative Fluid Intelligence (GFI)}$: the capacity to induce patterns, reason through constraints, and adapt to novel scenarios on the fly. To rigorously assess this capability, we introduce $\textbf{GENIUS}$ ($\textbf{GEN}$ Fluid $\textbf{I}$ntelligence Eval$\textbf{U}$ation $\textbf{S}$uite). We formalize $\textit{GFI}$ as a synthesis of three primitives. These include $\textit{Inducing Implicit Patterns}$ (e.g., inferring personalized visual preferences), $\textit{Executing Ad-hoc Constraints}$ (e.g., visualizing abstract metaphors), and $\textit{Adapting to Contextual Knowledge}$ (e.g., simulating counter-intuitive physics). Collectively, these primitives challenge models to solve problems grounded entirely in the immediate context. Our systematic evaluation of 12 representative models reveals significant performance deficits in these tasks. Crucially, our diagnostic analysis disentangles these failure modes. It demonstrates that deficits stem from limited context comprehension rather than insufficient intrinsic generative capability. To bridge this gap, we propose a training-free attention intervention strategy. Ultimately, $\textbf{GENIUS}$ establishes a rigorous standard for $\textit{GFI}$, guiding the field beyond knowledge utilization toward dynamic, general-purpose reasoning. Our dataset and code will be released at: $\href{https://github.com/arctanxarc/GENIUS}{https://github.com/arctanxarc/GENIUS}$.

---

### Data-Efficient Hierarchical Goal-Conditioned Reinforcement Learning via Normalizing Flows

Authors: Shaswat Garg, Matin Moezzi, Brandon Da Silva
Published: 2026-02-11
Categories: cs.RO, cs.AI, cs.LG
arXiv: https://arxiv.org/abs/2602.11142v1

Hierarchical goal-conditioned reinforcement learning (H-GCRL) provides a powerful framework for tackling complex, long-horizon tasks by decomposing them into structured subgoals. However, its practical adoption is hindered by poor data efficiency and limited policy expressivity, especially in offline or data-scarce regimes. In this work, Normalizing flow-based hierarchical implicit Q-learning (NF-HIQL), a novel framework that replaces unimodal gaussian policies with expressive normalizing flow policies at both the high- and low-levels of the hierarchy is introduced. This design enables tractable log-likelihood computation, efficient sampling, and the ability to model rich multimodal behaviors. New theoretical guarantees are derived, including explicit KL-divergence bounds for Real-valued non-volume preserving (RealNVP) policies and PAC-style sample efficiency results, showing that NF-HIQL preserves stability while improving generalization. Empirically, NF-HIQL is evaluted across diverse long-horizon tasks in locomotion, ball-dribbling, and multi-step manipulation from OGBench. NF-HIQL consistently outperforms prior goal-conditioned and hierarchical baselines, demonstrating superior robustness under limited data and highlighting the potential of flow-based architectures for scalable, data-efficient hierarchical reinforcement learning.

---

### LCIP: Loss-Controlled Inverse Projection of High-Dimensional Image Data

Authors: Yu Wang, Frederik L. Dennig, Michael Behrisch, Alexandru Telea
Published: 2026-02-11
Categories: cs.HC, cs.LG
arXiv: https://arxiv.org/abs/2602.11141v1

Projections (or dimensionality reduction) methods $P$ aim to map high-dimensional data to typically 2D scatterplots for visual exploration. Inverse projection methods $P^{-1}$ aim to map this 2D space to the data space to support tasks such as data augmentation, classifier analysis, and data imputation. Current $P^{-1}$ methods suffer from a fundamental limitation -- they can only generate a fixed surface-like structure in data space, which poorly covers the richness of this space. We address this by a new method that can `sweep' the data space under user control. Our method works generically for any $P$ technique and dataset, is controlled by two intuitive user-set parameters, and is simple to implement. We demonstrate it by an extensive application involving image manipulation for style transfer.

---

### TabICLv2: A better, faster, scalable, and open tabular foundation model

Authors: Jingang Qu, David Holzmüller, Gaël Varoquaux, Marine Le Morvan
Published: 2026-02-11
Categories: cs.LG
arXiv: https://arxiv.org/abs/2602.11139v2

Tabular foundation models, such as TabPFNv2 and TabICL, have recently dethroned gradient-boosted trees at the top of predictive benchmarks, demonstrating the value of in-context learning for tabular data. We introduce TabICLv2, a new state-of-the-art foundation model for regression and classification built on three pillars: (1) a novel synthetic data generation engine designed for high pretraining diversity; (2) various architectural innovations, including a new scalable softmax in attention improving generalization to larger datasets without prohibitive long-sequence pretraining; and (3) optimized pretraining protocols, notably replacing AdamW with the Muon optimizer. On the TabArena and TALENT benchmarks, TabICLv2 without any tuning surpasses the performance of the current state of the art, RealTabPFN-2.5 (hyperparameter-tuned, ensembled, and fine-tuned on real data). With only moderate pretraining compute, TabICLv2 generalizes effectively to million-scale datasets under 50 GB GPU memory while being markedly faster than RealTabPFN-2.5. We provide extensive ablation studies to quantify these contributions and foster open research by releasing code for inference, pretraining, and synthetic data generation at https://github.com/soda-inria/tabicl.

---

### TabICLv2: A better, faster, scalable, and open tabular foundation model

Authors: Jingang Qu, David Holzmüller, Gaël Varoquaux, Marine Le Morvan
Published: 2026-02-11
Categories: cs.LG
arXiv: https://arxiv.org/abs/2602.11139v1

Tabular foundation models, such as TabPFNv2 and TabICL, have recently dethroned gradient-boosted trees at the top of predictive benchmarks, demonstrating the value of in-context learning for tabular data. We introduce TabICLv2, a new state-of-the-art foundation model for regression and classification built on three pillars: (1) a novel synthetic data generation engine designed for high pretraining diversity; (2) various architectural innovations, including a new scalable softmax in attention improving generalization to larger datasets without prohibitive long-sequence pretraining; and (3) optimized pretraining protocols, notably replacing AdamW with the Muon optimizer. On the TabArena and TALENT benchmarks, TabICLv2 without any tuning surpasses the performance of the current state of the art, RealTabPFN-2.5 (hyperparameter-tuned, ensembled, and fine-tuned on real data). With only moderate pretraining compute, TabICLv2 generalizes effectively to million-scale datasets under 50GB GPU memory while being markedly faster than RealTabPFN-2.5. We provide extensive ablation studies to quantify these contributions and commit to open research by first releasing inference code and model weights at https://github.com/soda-inria/tabicl, with synthetic data engine and pretraining code to follow.

---

### Weight Decay Improves Language Model Plasticity

Authors: Tessa Han, Sebastian Bordt, Hanlin Zhang, Sham Kakade
Published: 2026-02-11
Categories: cs.LG, cs.AI, cs.CL
arXiv: https://arxiv.org/abs/2602.11137v2

Large language models are typically trained in two broad phases: pretraining to produce a base model, followed by further training to improve downstream performance. However, hyperparameter optimization and scaling laws are studied primarily from the perspective of the base model's validation loss, overlooking a crucial model property: downstream adaptability. In this work, we study pretraining from the perspective of model plasticity, that is, the ability of the base model to successfully adapt to downstream tasks upon additional training. We focus on the role of weight decay, a key regularization parameter during pretraining, and show through systematic experiments that larger weight decay increases the plasticity of the pretrained model, resulting in greater performance gains downstream after fine-tuning. This effect can lead to counterintuitive trade-offs where base models that perform worse after pretraining can perform better after further training. Further investigation of weight decay's mechanistic effects on model behavior reveals that it encourages linearly separable representations, regularizes attention matrices, and reduces overfitting on the training data. Together, these findings highlight the importance of pretrained model plasticity, the limits of using cross-entropy loss as the sole metric for hyperparameter optimization, and the multifaceted role that a single optimization hyperparameter plays in shaping model behavior.

---

### FormalJudge: A Neuro-Symbolic Paradigm for Agentic Oversight

Authors: Jiayi Zhou, Yang Sheng, Hantao Lou, Yaodong Yang, Jie Fu
Published: 2026-02-11
Categories: cs.AI
arXiv: https://arxiv.org/abs/2602.11136v2

As LLM-based agents increasingly operate in high-stakes domains with real-world consequences, ensuring their behavioral safety becomes paramount. The dominant oversight paradigm, LLM-as-a-Judge, faces a fundamental dilemma: how can probabilistic systems reliably supervise other probabilistic systems without inheriting their failure modes? We argue that formal verification offers a principled escape from this dilemma, yet its adoption has been hindered by a critical bottleneck: the translation from natural language requirements to formal specifications. This paper bridges this gap by proposing , a neuro-symbolic framework that employs a bidirectional Formal-of-Thought architecture: LLMs serve as specification compilers that top-down decompose high-level human intent into atomic, verifiable constraints, then bottom-up prove compliance using Dafny specifications and Z3 Satisfiability modulo theories solving, which produces mathematical guarantees rather than probabilistic scores. We validate across three benchmarks spanning behavioral safety, multi-domain constraint adherence, and agentic upward deception detection. Experiments on 7 agent models demonstrate that achieves an average improvement of 16.6% over LLM-as-a-Judge baselines, enables weak-to-strong generalization where a 7B judge achieves over 90% accuracy detecting deception from 72B agents, and provides near-linear safety improvement through iterative refinement.

---

### Just on Time: Token-Level Early Stopping for Diffusion Language Models

Authors: Zakhar Kohut, Severyn Shykula, Mykola Vysotskyi, Serhii Dmytryshyn, Dmytro Khamula, Michal Zakrzewski, Damian Rynczak, Jacek Małecki, Taras Rumezhak, Volodymyr Karpiv
Published: 2026-02-11
Categories: cs.LG, cs.CL
arXiv: https://arxiv.org/abs/2602.11133v2

Diffusion language models generate text through iterative refinement, a process that is often computationally inefficient because many tokens reach stability long before the final denoising step. We introduce a training-free, token-level early stopping approach that identifies convergence independently at each position. Our method leverages lightweight signals derived from the model's predictions and local context to dynamically determine when individual tokens can be finalized. This yields adaptive per-token freezing without task-specific fine-tuning, substantially reducing the total number of diffusion steps required. Across diverse benchmarks, spanning mathematical reasoning, general question answering, and scientific understanding, our approach achieves substantial efficiency gains while preserving generation quality.

---

### Meltdown: Circuits and Bifurcations in Point-Cloud-Conditioned 3D Diffusion Transformers

Authors: Maximilian Plattner, Fabian Paischer, Johannes Brandstetter, Arturs Berzins
Published: 2026-02-11
Categories: cs.LG, cs.CV
arXiv: https://arxiv.org/abs/2602.11130v2

Sparse point clouds are a common input modality for 3D surface reconstruction, including in safety-critical settings such as surgical navigation and autonomous perception. Recent point-cloud-conditioned 3D diffusion transformers achieve state-of-the-art results in this regime by leveraging learned priors. We show that these models can fail catastrophically under realistic input variation, and present a mechanistic case study of why. We identify a failure mode we call Meltdown: tiny on-surface perturbations to a sparse input point cloud can fracture the reconstructed output into hundreds of disconnected pieces. Adversarial search recovers Meltdown in 89.9-100% of shapes across the two open-weight state-of-the-art architectures we study (WaLa, Make-a-Shape) on real-world datasets (GSO, SimJEB) and under both DDPM and DDIM sampling. We trace Meltdown along the forward pass: it is governed by how uniformly the points are distributed on the surface, faithfully transduced through the point-cloud encoder, and committed by a single early-denoising cross-attention write in the diffusion backbone. Diffusion-trajectory ensembles exhibit symmetry-breaking near this commit step, consistent with a bifurcation of the reverse process. Through a suite of matched-magnitude controls, we show that the variable on which the model commits is directional, concentrated in a low-rank subspace of the write's perturbation drift. Motivated by this finding, we introduce PowerRemap, a test-time control that reshapes the singular spectrum of the localized write to suppress this drift, with rescue rates of 98.3% on WaLa and 84.6% on Make-a-Shape. Together, these results link a circuit-level cross-attention mechanism to a trajectory-level account of the failure, demonstrating how mechanistic analysis can explain and guide behavior in conditional diffusion transformers.

---

### Asymmetric Prompt Weighting for Reinforcement Learning with Verifiable Rewards

Authors: Reinhard Heckel, Mahdi Soltanolkotabi, Christos Thramboulidis
Published: 2026-02-11
Categories: cs.LG
arXiv: https://arxiv.org/abs/2602.11128v1

Reinforcement learning with verifiable rewards has driven recent advances in LLM post-training, in particular for reasoning. Policy optimization algorithms generate a number of responses for a given prompt and then effectively weight the corresponding gradients depending on the rewards. The most popular algorithms including GRPO, DAPO, and RLOO focus on ambiguous prompts, i.e., prompts with intermediate success probability, while downgrading gradients with very easy and very hard prompts. In this paper, we consider asymmetric prompt weightings that assign higher weights to prompts with low, or even zero, empirical success probability. We find that asymmetric weighting particularly benefits from-scratch RL (as in R1-Zero), where training traverses a wide accuracy range, and less so in post-SFT RL where the model already starts at high accuracy. We also provide theory that characterizes prompt weights which minimize the time needed to raise success probability from an initial level to a target accuracy under a fixed update budget. In low-success regimes, where informative responses are rare and response cost dominates, these optimal weights become asymmetric, upweighting low success probabilities and thereby accelerating effective-time convergence.

---

### The Offline-Frontier Shift: Diagnosing Distributional Limits in Generative Multi-Objective Optimization

Authors: Stephanie Holly, Alexandru-Ciprian Zăvoianu, Siegfried Silber, Sepp Hochreiter, Werner Zellinger
Published: 2026-02-11
Categories: cs.LG
arXiv: https://arxiv.org/abs/2602.11126v2

Offline multi-objective optimization (MOO) aims to recover Pareto-optimal designs given a finite, static dataset. Recent generative approaches, including diffusion models, show strong performance under hypervolume, yet their behavior under other established MOO metrics is less understood. We show that generative methods systematically underperform evolutionary alternatives with respect to other metrics, such as generational distance. We relate this failure mode to the offline-frontier shift, i.e., the displacement of the offline dataset from the Pareto front, which acts as a fundamental limitation in offline MOO. We argue that overcoming this limitation requires out-of-distribution sampling in objective space (via an integral probability metric) and empirically observe that generative methods remain conservatively close to the offline objective distribution. Our results position offline MOO as a distribution-shift--limited problem and provide a diagnostic lens for understanding when and why generative optimization methods fail.

---

### PhyCritic: Multimodal Critic Models for Physical AI

Authors: Tianyi Xiong, Shihao Wang, Guilin Liu, Yi Dong, Ming Li, Heng Huang, Jan Kautz, Zhiding Yu
Published: 2026-02-11
Categories: cs.CV
arXiv: https://arxiv.org/abs/2602.11124v1

With the rapid development of large multimodal models, reliable judge and critic models have become essential for open-ended evaluation and preference alignment, providing pairwise preferences, numerical scores, and explanatory justifications for assessing model-generated responses. However, existing critics are primarily trained in general visual domains such as captioning or image question answering, leaving physical AI tasks involving perception, causal reasoning, and planning largely underexplored. We introduce PhyCritic, a multimodal critic model optimized for physical AI through a two-stage RLVR pipeline: a physical skill warmup stage that enhances physically oriented perception and reasoning, followed by self-referential critic finetuning, where the critic generates its own prediction as an internal reference before judging candidate responses, improving judgment stability and physical correctness. Across both physical and general-purpose multimodal judge benchmarks, PhyCritic achieves strong performance gains over open-source baselines and, when applied as a policy model, further improves perception and reasoning in physically grounded tasks.

---

### From Natural Language to Materials Discovery:The Materials Knowledge Navigation Agent

Authors: Genmao Zhuang, Amir Barati Farimani
Published: 2026-02-11
Categories: cs.LG, cond-mat.mtrl-sci
arXiv: https://arxiv.org/abs/2602.11123v1

Accelerating the discovery of high-performance materials remains a central challenge across energy, electronics, and aerospace technologies, where traditional workflows depend heavily on expert intuition and computationally expensive simulations. Here we introduce the Materials Knowledge Navigation Agent (MKNA), a language-driven system that translates natural-language scientific intent into executable actions for database retrieval, property prediction, structure generation, and stability evaluation. Beyond automating tool invocation, MKNA autonomously extracts quantitative thresholds and chemically meaningful design motifs from literature and database evidence, enabling data-grounded hypothesis formation. Applied to the search for high-Debye-temperature ceramics, the agent identifies a literature-supported screening criterion (Theta_D > 800 K), rediscovers canonical ultra-stiff materials such as diamond, SiC, SiN, and BeO, and proposes thermodynamically stable, previously unreported Be-C-rich compounds that populate the sparsely explored 1500-1700 K regime. These results demonstrate that MKNA not only finds stable candidates but also reconstructs interpretable design heuristics, establishing a generalizable platform for autonomous, language-guided materials exploration.

---

### A Doubly Robust Machine Learning Approach for Disentangling Treatment Effect Heterogeneity with Functional Outcomes

Authors: Filippo Salmaso, Lorenzo Testa, Francesca Chiaromonte
Published: 2026-02-11
Categories: stat.ME, stat.ML
arXiv: https://arxiv.org/abs/2602.11118v1

Causal inference is paramount for understanding the effects of interventions, yet extracting personalized insights from increasingly complex data remains a significant challenge for modern machine learning. This is the case, in particular, when considering functional outcomes observed over a continuous domain (e.g., time, or space). Estimation of heterogeneous treatment effects, known as CATE, has emerged as a crucial tool for personalized decision-making, but existing meta-learning frameworks are largely limited to scalar outcomes, failing to provide satisfying results in scientific applications that leverage the rich, continuous information encoded in functional data. Here, we introduce FOCaL (Functional Outcome Causal Learning), a novel, doubly robust meta-learner specifically engineered to estimate a functional heterogeneous treatment effect (F-CATE). FOCaL integrates advanced functional regression techniques for both outcome modeling and functional pseudo-outcome reconstruction, thereby enabling the direct and robust estimation of F-CATE. We provide a rigorous theoretical derivation of FOCaL, demonstrate its performance and robustness compared to existing non-robust functional methods through comprehensive simulation studies, and illustrate its practical utility on diverse real-world functional datasets. FOCaL advances the capabilities of machine intelligence to infer nuanced, individualized causal effects from complex data, paving the way for more precise and trustworthy AI systems in personalized medicine, adaptive policy design, and fundamental scientific discovery.

---

### HairWeaver: Few-Shot Photorealistic Hair Motion Synthesis with Sim-to-Real Guided Video Diffusion

Authors: Di Chang, Ji Hou, Aljaz Bozic, Assaf Neuberger, Felix Juefei-Xu, Olivier Maury, Gene Wei-Chin Lin, Tuur Stuyck, Doug Roble, Mohammad Soleymani, Stephane Grabli
Published: 2026-02-11
Categories: cs.CV
arXiv: https://arxiv.org/abs/2602.11117v2

We present HairWeaver, a diffusion-based pipeline that animates a single human image with realistic and expressive hair dynamics. While existing methods successfully control body pose, they lack specific control over hair, and as a result, fail to capture the intricate hair motions, resulting in stiff and unrealistic animations. HairWeaver overcomes this limitation using two specialized modules: a Motion-Context-LoRA to integrate motion conditions and a Style-Alignment-LoRA to preserve the subject's photoreal appearance across different data domains. These lightweight components are designed to guide a video diffusion backbone while maintaining its core generative capabilities. By training on a specialized dataset of dynamic human motion generated from a CG simulator, HairWeaver affords fine control over hair motion and ultimately learns to produce highly realistic hair that responds naturally to movement. Comprehensive evaluations demonstrate that our approach sets a new state of the art, producing lifelike human hair animations with dynamic details.

---

### Learning to Compose for Cross-domain Agentic Workflow Generation

Authors: Jialiang Wang, Shengxiang Xu, Hanmo Liu, Jiachuan Wang, Yuyu Luo, Shimin Di, Min-Ling Zhang, Lei Chen
Published: 2026-02-11
Categories: cs.MA, cs.AI, cs.LG, cs.SE
arXiv: https://arxiv.org/abs/2602.11114v1

Automatically generating agentic workflows -- executable operator graphs or codes that orchestrate reasoning, verification, and repair -- has become a practical way to solve complex tasks beyond what single-pass LLM generation can reliably handle. Yet what constitutes a good workflow depends heavily on the task distribution and the available operators. Under domain shift, current systems typically rely on iterative workflow refinement to discover a feasible workflow from a large workflow space, incurring high iteration costs and yielding unstable, domain-specific behavior. In response, we internalize a decompose-recompose-decide mechanism into an open-source LLM for cross-domain workflow generation. To decompose, we learn a compact set of reusable workflow capabilities across diverse domains. To recompose, we map each input task to a sparse composition over these bases to generate a task-specific workflow in a single pass. To decide, we attribute the success or failure of workflow generation to counterfactual contributions from learned capabilities, thereby capturing which capabilities actually drive success by their marginal effects. Across stringent multi-domain, cross-domain, and unseen-domain evaluations, our 1-pass generator surpasses SOTA refinement baselines that consume 20 iterations, while substantially reducing generation latency and cost.

---

### Renet: Principled and Efficient Relaxation for the Elastic Net via Dynamic Objective Selection

Authors: Albert Dorador
Published: 2026-02-11
Categories: stat.ME, cs.LG, stat.ML
arXiv: https://arxiv.org/abs/2602.11107v1

We introduce Renet, a principled generalization of the Relaxed Lasso to the Elastic Net family of estimators. While, on the one hand, $\ell_1$-regularization is a standard tool for variable selection in high-dimensional regimes and, on the other hand, the $\ell_2$ penalty provides stability and solution uniqueness through strict convexity, the standard Elastic Net nevertheless suffers from shrinkage bias that frequently yields suboptimal prediction accuracy. We propose to address this limitation through a framework called \textit{relaxation}. Existing relaxation implementations rely on naive linear interpolations of penalized and unpenalized solutions, which ignore the non-linear geometry that characterizes the entire regularization path and risk violating the Karush-Kuhn-Tucker conditions. Renet addresses these limitations by enforcing sign consistency through an adaptive relaxation procedure that dynamically dispatches between convex blending and efficient sub-path refitting. Furthermore, we identify and formalize a unique synergy between relaxation and the ``One-Standard-Error'' rule: relaxation serves as a robust debiasing mechanism, allowing practitioners to leverage the parsimony of the 1-SE rule without the traditional loss in predictive fidelity. Our theoretical framework incorporates automated stability safeguards for ultra-high dimensional regimes and is supported by a comprehensive benchmarking suite across 20 synthetic and real-world datasets, demonstrating that Renet consistently outperforms the standard Elastic Net and provides a more robust alternative to the Adaptive Elastic Net in high-dimensional, low signal-to-noise ratio and high-multicollinearity regimes. By leveraging an adaptive solver backend, Renet delivers these statistical gains while offering a computational profile that remains competitive with state-of-the-art coordinate descent implementations.

---

### TEGRA: Text Encoding With Graph and Retrieval Augmentation for Misinformation Detection

Authors: Géraud Faye, Wassila Ouerdane, Guillaume Gadek, Sylvain Gatepaille, Céline Hudelot
Published: 2026-02-11
Categories: cs.CL
arXiv: https://arxiv.org/abs/2602.11106v2

Misinformation detection is a critical task that can benefit significantly from the integration of external knowledge, much like manual fact-checking. In this work, we propose a novel method for representing textual documents that facilitates the incorporation of information from a knowledge base. Our approach, Text Encoding with Graph (TEG), processes documents by extracting structured information in the form of a graph and encoding both the text and the graph for classification purposes. Through extensive experiments, we demonstrate that this hybrid representation enhances misinformation detection performance compared to using language models alone. Furthermore, we introduce TEGRA, an extension of our framework that integrates domain-specific knowledge, further enhancing classification accuracy in most cases.

---

### FastFlow: Accelerating The Generative Flow Matching Models with Bandit Inference

Authors: Divya Jyoti Bajpai, Dhruv Bhardwaj, Soumya Roy, Tejas Duseja, Harsh Agarwal, Aashay Sandansing, Manjesh Kumar Hanawal
Published: 2026-02-11
Categories: cs.CV
arXiv: https://arxiv.org/abs/2602.11105v1

Flow-matching models deliver state-of-the-art fidelity in image and video generation, but the inherent sequential denoising process renders them slower. Existing acceleration methods like distillation, trajectory truncation, and consistency approaches are static, require retraining, and often fail to generalize across tasks. We propose FastFlow, a plug-and-play adaptive inference framework that accelerates generation in flow matching models. FastFlow identifies denoising steps that produce only minor adjustments to the denoising path and approximates them without using the full neural network models used for velocity predictions. The approximation utilizes finite-difference velocity estimates from prior predictions to efficiently extrapolate future states, enabling faster advancements along the denoising path at zero compute cost. This enables skipping computation at intermediary steps. We model the decision of how many steps to safely skip before requiring a full model computation as a multi-armed bandit problem. The bandit learns the optimal skips to balance speed with performance. FastFlow integrates seamlessly with existing pipelines and generalizes across image generation, video generation, and editing tasks. Experiments demonstrate a speedup of over 2.6x while maintaining high-quality outputs. The source code for this work can be found at https://github.com/Div290/FastFlow.

---

### GameDevBench: Evaluating Agentic Capabilities Through Game Development

Authors: Wayne Chi, Yixiong Fang, Arnav Yayavaram, Siddharth Yayavaram, Seth Karten, Qiuhong Anna Wei, Runkun Chen, Alexander Wang, Valerie Chen, Ameet Talwalkar, Chris Donahue
Published: 2026-02-11
Categories: cs.AI, cs.CL, cs.SE
arXiv: https://arxiv.org/abs/2602.11103v2

Despite rapid progress on coding agents, progress on their multimodal counterparts has lagged behind. A key challenge is the scarcity of evaluation testbeds that combine the complexity of software development with the need for deep multimodal understanding. In game development, agents must navigate large, dense codebases while manipulating intrinsically multimodal assets such as shaders, sprites, and animations within a visual game scene. We present GameDevBench, the first benchmark for evaluating agents on game development tasks. GameDevBench consists of 333 tasks derived from web and video tutorials. Tasks require significant multimodal understanding and are complex: the average solution requires over three times the lines of code and file changes compared to prior software development benchmarks. Agents struggle with game development, with the best agent and method solving only 53.8% of tasks. We find a strong correlation between perceived task difficulty and multimodal complexity, with average success rate dropping from 51.4% on gameplay-oriented tasks to 33.0% on 2D graphics tasks. To improve multimodal capability, we introduce two simple image- and video-based feedback mechanisms for agents. Despite their simplicity, these methods consistently improve performance, increasing GPT-5.4's performance from 41.1% to 52.0% when given visual feedback.

---

### Statistical Learning Analysis of Physics-Informed Neural Networks

Authors: David A. Barajas-Solano
Published: 2026-02-11
Categories: cs.LG, physics.comp-ph
arXiv: https://arxiv.org/abs/2602.11097v1

We study the training and performance of physics-informed learning for initial and boundary value problems (IBVP) with physics-informed neural networks (PINNs) from a statistical learning perspective. Specifically, we restrict ourselves to parameterizations with hard initial and boundary condition constraints and reformulate the problem of estimating PINN parameters as a statistical learning problem. From this perspective, the physics penalty on the IBVP residuals can be better understood not as a regularizing term bus as an infinite source of indirect data, and the learning process as fitting the PINN distribution of residuals $p(y \mid x, t, w) q(x, t) $ to the true data-generating distribution $δ(0) q(x, t)$ by minimizing the Kullback-Leibler divergence between the true and PINN distributions. Furthermore, this analysis show that physics-informed learning with PINNs is a singular learning problem, and we employ singular learning theory tools, namely the so-called Local Learning Coefficient (Lau et al., 2025) to analyze the estimates of PINN parameters obtained via stochastic optimization for a heat equation IBVP. Finally, we discuss implications of this analysis on the quantification of predictive uncertainty of PINNs and the extrapolation capacity of PINNs.

---

### Safety Recovery in Reasoning Models Is Only a Few Early Steering Steps Away

Authors: Soumya Suvra Ghosal, Souradip Chakraborty, Vaibhav Singh, Furong Huang, Dinesh Manocha, Amrit Singh Bedi
Published: 2026-02-11
Categories: cs.CL, cs.AI
arXiv: https://arxiv.org/abs/2602.11096v1

Reinforcement learning (RL) based post-training for explicit chain-of-thought (e.g., GRPO) improves the reasoning ability of multimodal large-scale reasoning models (MLRMs). But recent evidence shows that it can simultaneously degrade safety alignment and increase jailbreak success rates. We propose SafeThink, a lightweight inference-time defense that treats safety recovery as a satisficing constraint rather than a maximization objective. SafeThink monitors the evolving reasoning trace with a safety reward model and conditionally injects an optimized short corrective prefix ("Wait, think safely") only when the safety threshold is violated. In our evaluations across six open-source MLRMs and four jailbreak benchmarks (JailbreakV-28K, Hades, FigStep, and MM-SafetyBench), SafeThink reduces attack success rates by 30-60% (e.g., LlamaV-o1: 63.33% to 5.74% on JailbreakV-28K, R1-Onevision: 69.07% to 5.65% on Hades) while preserving reasoning performance (MathVista accuracy: 65.20% to 65.00%). A key empirical finding from our experiments is that safety recovery is often only a few steering steps away: intervening in the first 1-3 reasoning steps typically suffices to redirect the full generation toward safe completions.

---

### MerLin: A Discovery Engine for Photonic and Hybrid Quantum Machine Learning

Authors: Cassandre Notton, Benjamin Stott, Philippe Schoeb, Anthony Walsh, Grégoire Leboucher, Vincent Espitalier, Vassilis Apostolou, Louis-Félix Vigneux, Alexia Salavrakos, Jean Senellart
Published: 2026-02-11
Categories: cs.LG, cs.PL, quant-ph
arXiv: https://arxiv.org/abs/2602.11092v2

Identifying where quantum models may offer practical benefits in near term quantum machine learning (QML) requires moving beyond isolated algorithmic proposals toward systematic and empirical exploration across models, datasets, and hardware constraints. We introduce MerLin, an open-source framework designed as a discovery engine for photonic and hybrid quantum machine learning. MerLin integrates optimized strong simulation of linear optical circuits into standard PyTorch and scikit learn workflows, enabling end-to-end differentiable training of quantum layers. MerLin is designed around systematic benchmarking and reproducibility. As an initial contribution, we reproduce eighteen state-of-the-art photonic and hybrid QML works spanning kernel methods, reservoir computing, convolutional and recurrent architectures, generative models, and modern training paradigms. These reproductions are released as reusable, modular experiments that can be directly extended and adapted, establishing a shared experimental baseline consistent with empirical benchmarking methodologies widely adopted in modern artificial intelligence. By embedding photonic quantum models within established machine learning ecosystems, MerLin allows practitioners to leverage existing tooling for ablation studies, cross-modality comparisons, and hybrid classical-quantum workflows. The framework already implements hardware-aware features, allowing tests on available quantum hardware while enabling exploration beyond its current capabilities, positioning MerLin as a forward-looking co-design tool linking algorithms, benchmarks, and hardware.

---

### Can Large Language Models Make Everyone Happy?

Authors: Usman Naseem, Gautam Siddharth Kashyap, Ebad Shabbir, Sushant Kumar Ray, Abdullah Mohammad, Rafiq Ali
Published: 2026-02-11
Categories: cs.CL
arXiv: https://arxiv.org/abs/2602.11091v1

Misalignment in Large Language Models (LLMs) refers to the failure to simultaneously satisfy safety, value, and cultural dimensions, leading to behaviors that diverge from human expectations in real-world settings where these dimensions must co-occur. Existing benchmarks, such as SAFETUNEBED (safety-centric), VALUEBENCH (value-centric), and WORLDVIEW-BENCH (culture-centric), primarily evaluate these dimensions in isolation and therefore provide limited insight into their interactions and trade-offs. More recent efforts, including MIB and INTERPRETABILITY BENCHMARK-based on mechanistic interpretability, offer valuable perspectives on model failures; however, they remain insufficient for systematically characterizing cross-dimensional trade-offs. To address these gaps, we introduce MisAlign-Profile, a unified benchmark for measuring misalignment trade-offs inspired by mechanistic profiling. First, we construct MISALIGNTRADE, an English misaligned-aligned dataset across 112 normative domains taxonomies, including 14 safety, 56 value, and 42 cultural domains. In addition to domain labels, each prompt is classified with one of three orthogonal semantic types-object, attribute, or relations misalignment-using Gemma-2-9B-it and expanded via Qwen3-30B-A3B-Instruct-2507 with SimHash-based fingerprinting to avoid deduplication. Each prompt is paired with misaligned and aligned responses through two-stage rejection sampling to ensure quality. Second, we benchmark general-purpose, fine-tuned, and open-weight LLMs on MISALIGNTRADE-revealing 12%-34% misalignment trade-offs across dimensions.

---

### Direct Learning of Calibration-Aware Uncertainty for Neural PDE Surrogates

Authors: Carlos Stein Brito
Published: 2026-02-11
Categories: cs.LG, cs.AI, cs.CE, stat.CO
arXiv: https://arxiv.org/abs/2602.11090v1

Neural PDE surrogates are often deployed in data-limited or partially observed regimes where downstream decisions depend on calibrated uncertainty in addition to low prediction error. Existing approaches obtain uncertainty through ensemble replication, fixed stochastic noise such as dropout, or post hoc calibration. Cross-regularized uncertainty learns uncertainty parameters during training using gradients routed through a held-out regularization split. The predictor is optimized on the training split for fit, while low-dimensional uncertainty controls are optimized on the regularization split to reduce train-test mismatch, yielding regime-adaptive uncertainty without per-regime noise tuning. The framework can learn continuous noise levels at the output head, within hidden features, or within operator-specific components such as spectral modes. We instantiate the approach in Fourier Neural Operators and evaluate on APEBench sweeps over observed fraction and training-set size. Across these sweeps, the learned predictive distributions are better calibrated on held-out splits and the resulting uncertainty fields concentrate in high-error regions in one-step spatial diagnostics.

---

### DataChef: Cooking Up Optimal Data Recipes for LLM Adaptation via Reinforcement Learning

Authors: Yicheng Chen, Zerun Ma, Xinchen Xie, Yining Li, Kai Chen
Published: 2026-02-11
Categories: cs.CL, cs.AI
arXiv: https://arxiv.org/abs/2602.11089v2

In the current landscape of Large Language Models (LLMs), the curation of large-scale, high-quality training data is a primary driver of model performance. A key lever is the \emph{data recipe}, which comprises a data processing pipeline to transform raw sources into training corpora. Despite the growing use of LLMs to automate individual data processing steps, such as data synthesis and filtering, the overall design of data recipes remains largely manual and labor-intensive, requiring substantial human expertise and iteration. To bridge this gap, we formulate \emph{end-to-end data recipe generation} for LLM adaptation. Given a target benchmark and a pool of available data sources, a model is required to output a complete data recipe that adapts a base LLM to the target task. We present DataChef-32B, which performs online reinforcement learning using a proxy reward that predicts downstream performance for candidate recipes. Across six held-out tasks, DataChef-32B produces recipes that yield performance comparable to those curated by human experts. Notably, the recipe from DataChef-32B adapts Qwen3-1.7B-Base to the math domain, achieving 66.7 on AIME'25 and surpassing the official post-training checkpoint (Qwen3-1.7B). This work sheds new light on automating LLM training and developing self-evolving AI systems.

---

### General Flexible $f$-divergence for Challenging Offline RL Datasets with Low Stochasticity and Diverse Behavior Policies

Authors: Jianxun Wang, Grant C. Forbes, Leonardo Villalobos-Arias, David L. Roberts
Published: 2026-02-11
Categories: cs.LG, cs.AI
arXiv: https://arxiv.org/abs/2602.11087v1

Offline RL algorithms aim to improve upon the behavior policy that produces the collected data while constraining the learned policy to be within the support of the dataset. However, practical offline datasets often contain examples with little diversity or limited exploration of the environment, and from multiple behavior policies with diverse expertise levels. Limited exploration can impair the offline RL algorithm's ability to estimate \textit{Q} or \textit{V} values, while constraining towards diverse behavior policies can be overly conservative. Such datasets call for a balance between the RL objective and behavior policy constraints. We first identify the connection between $f$-divergence and optimization constraint on the Bellman residual through a more general Linear Programming form for RL and the convex conjugate. Following this, we introduce the general flexible function formulation for the $f$-divergence to incorporate an adaptive constraint on algorithms' learning objectives based on the offline training dataset. Results from experiments on the MuJoCo, Fetch, and AdroitHand environments show the correctness of the proposed LP form and the potential of the flexible $f$-divergence in improving performance for learning from a challenging dataset when applied to a compatible constrained optimization algorithm.

---

### First International StepUP Competition for Biometric Footstep Recognition: Methods, Results and Remaining Challenges

Authors: Robyn Larracy, Eve MacDonald, Angkoon Phinyomark, Saeid Rezaei, Mahdi Laghaei, Ali Hajighasem, Aaron Tabor, Erik Scheme
Published: 2026-02-11
Categories: cs.CV, cs.LG
arXiv: https://arxiv.org/abs/2602.11086v1

Biometric footstep recognition, based on a person's unique pressure patterns under their feet during walking, is an emerging field with growing applications in security and safety. However, progress in this area has been limited by the lack of large, diverse datasets necessary to address critical challenges such as generalization to new users and robustness to shifts in factors like footwear or walking speed. The recent release of the UNB StepUP-P150 dataset, the largest and most comprehensive collection of high-resolution footstep pressure recordings to date, opens new opportunities for addressing these challenges through deep learning. To mark this milestone, the First International StepUP Competition for Biometric Footstep Recognition was launched. Competitors were tasked with developing robust recognition models using the StepUP-P150 dataset that were then evaluated on a separate, dedicated test set designed to assess verification performance under challenging variations, given limited and relatively homogeneous reference data. The competition attracted global participation, with 23 registered teams from academia and industry. The top-performing team, Saeid_UCC, achieved the best equal error rate (EER) of 10.77% using a generative reward machine (GRM) optimization strategy. Overall, the competition showcased strong solutions, but persistent challenges in generalizing to unfamiliar footwear highlight a critical area for future work.

---

### GRASP: group-Shapley feature selection for patients

Authors: Yuheng Luo, Shuyan Li, Zhong Cao
Published: 2026-02-11
Categories: cs.LG, cs.AI
arXiv: https://arxiv.org/abs/2602.11084v2

Feature selection remains a major challenge in medical prediction, where existing approaches such as LASSO often lack robustness and interpretability. We introduce GRASP, a novel framework that couples Shapley value driven attribution with group $L_{21}$ regularization to extract compact and non-redundant feature sets. GRASP first distills group level importance scores from a pretrained tree model via SHAP, then enforces structured sparsity through group $L_{21}$ regularized logistic regression, yielding stable and interpretable selections. Extensive comparisons with LASSO, SHAP, and deep learning based methods show that GRASP consistently delivers comparable or superior predictive accuracy, while identifying fewer, less redundant, and more stable features.

---

### Token-Efficient Change Detection in LLM APIs

Authors: Timothée Chauvin, Clément Lalanne, Erwan Le Merrer, Jean-Michel Loubes, François Taïani, Gilles Tredan
Published: 2026-02-11
Categories: cs.LG, cs.CR
arXiv: https://arxiv.org/abs/2602.11083v4

Remote change detection in LLMs is a difficult problem. Existing methods are either too expensive for deployment at scale, or require initial white-box access to model weights or grey-box access to log probabilities. We aim to achieve both low cost and strict black-box operation, observing only output tokens. Our approach hinges on specific inputs we call Border Inputs, for which there exists more than one output top token. From a statistical perspective, optimal change detection depends on the model's Jacobian and the Fisher information of the output distribution. Analyzing these quantities in low-temperature regimes shows that border inputs enable powerful change detection tests. Building on this insight, we propose the Black-Box Border Input Tracking (B3IT) scheme. Extensive in-vivo and in-vitro experiments show that border inputs are easily found for non-reasoning tested endpoints, and achieve performance on par with the best available grey-box approaches. B3IT reduces costs by $30\times$ compared to existing methods, while operating in a strict black-box setting.

---

### SteuerLLM: Local specialized large language model for German tax law analysis

Authors: Sebastian Wind, Jeta Sopa, Laurin Schmid, Quirin Jackl, Sebastian Kiefer, Fei Wu, Martin Mayr, Harald Köstler, Gerhard Wellein, Andreas Maier, Soroosh Tayebi Arasteh
Published: 2026-02-11
Categories: cs.CL, cs.AI, cs.LG
arXiv: https://arxiv.org/abs/2602.11081v1

Large language models (LLMs) demonstrate strong general reasoning and language understanding, yet their performance degrades in domains governed by strict formal rules, precise terminology, and legally binding structure. Tax law exemplifies these challenges, as correct answers require exact statutory citation, structured legal argumentation, and numerical accuracy under rigid grading schemes. We algorithmically generate SteuerEx, the first open benchmark derived from authentic German university tax law examinations. SteuerEx comprises 115 expert-validated examination questions spanning six core tax law domains and multiple academic levels, and employs a statement-level, partial-credit evaluation framework that closely mirrors real examination practice. We further present SteuerLLM, a domain-adapted LLM for German tax law trained on a large-scale synthetic dataset generated from authentic examination material using a controlled retrieval-augmented pipeline. SteuerLLM (28B parameters) consistently outperforms general-purpose instruction-tuned models of comparable size and, in several cases, substantially larger systems, demonstrating that domain-specific data and architectural adaptation are more decisive than parameter scale for performance on realistic legal reasoning tasks. All benchmark data, training datasets, model weights, and evaluation code are released openly to support reproducible research in domain-specific legal artificial intelligence. A web-based demo of SteuerLLM is available at https://steuerllm.i5.ai.fau.de.

---

### Probe-Based Data Attribution: Discovering and Mitigating Undesirable Behaviors in LLM Post-Training

Authors: Frank Xiao, Santiago Aranguri
Published: 2026-02-11
Categories: cs.LG, cs.AI
arXiv: https://arxiv.org/abs/2602.11079v3

We propose probe-based data attribution, a method that traces behavioral changes in post-trained language models to responsible training datapoints. By computing activation-difference vectors for both test prompts and preference pairs and ranking by cosine similarity, we identify datapoints that cause specific behaviors and validate these attributions causally by retraining with modified data. Clustering behavior-datapoint similarity matrices also enables unsupervised discovery of emergent behaviors. Applying this to OLMo 2's production DPO training, we surfaced distractor-triggered compliance: a harmful behavior where the model complies with dangerous requests when benign formatting instructions are appended. Filtering top-ranked datapoints reduces this behavior by 63% while switching their labels achieves 78%. Our method outperforms gradient-based attribution and LLM-judge baselines while being over 10 times cheaper than both. This in-the-wild model organism - emerging from contaminated preference data rather than deliberate injection - provides a realistic benchmark for safety techniques.

---

### Interpretable Attention-Based Multi-Agent PPO for Latency Spike Resolution in 6G RAN Slicing

Authors: Kavan Fatehi, Mostafa Rahmani Ghourtani, Amir Sonee, Poonam Yadav, Alessandra M Russo, Hamed Ahmadi, Radu Calinescu
Published: 2026-02-11
Categories: eess.SY, cs.AI, eess.SP
arXiv: https://arxiv.org/abs/2602.11076v1

Sixth-generation (6G) radio access networks (RANs) must enforce strict service-level agreements (SLAs) for heterogeneous slices, yet sudden latency spikes remain difficult to diagnose and resolve with conventional deep reinforcement learning (DRL) or explainable RL (XRL). We propose \emph{Attention-Enhanced Multi-Agent Proximal Policy Optimization (AE-MAPPO)}, which integrates six specialized attention mechanisms into multi-agent slice control and surfaces them as zero-cost, faithful explanations. The framework operates across O-RAN timescales with a three-phase strategy: predictive, reactive, and inter-slice optimization. A URLLC case study shows AE-MAPPO resolves a latency spike in $18$ms, restores latency to $0.98$ms with $99.9999\%$ reliability, and reduces troubleshooting time by $93\%$ while maintaining eMBB and mMTC continuity. These results confirm AE-MAPPO's ability to combine SLA compliance with inherent interpretability, enabling trustworthy and real-time automation for 6G RAN slicing.

---

### Chatting with Images for Introspective Visual Thinking

Authors: Junfei Wu, Jian Guan, Qiang Liu, Shu Wu, Liang Wang, Wei Wu, Tieniu Tan
Published: 2026-02-11
Categories: cs.CV, cs.AI, cs.CL
arXiv: https://arxiv.org/abs/2602.11073v2

Current large vision-language models (LVLMs) typically rely on text-only reasoning based on a single-pass visual encoding, which often leads to loss of fine-grained visual information. Recently the proposal of ''thinking with images'' attempts to alleviate this limitation by manipulating images via external tools or code; however, the resulting visual states are often insufficiently grounded in linguistic semantics, impairing effective cross-modal alignment - particularly when visual semantics or geometric relationships must be reasoned over across distant regions or multiple images. To address these challenges, we propose ''chatting with images'', a new framework that reframes visual manipulation as language-guided feature modulation. Under the guidance of expressive language prompts, the model dynamically performs joint re-encoding over multiple image regions, enabling tighter coupling between linguistic reasoning and visual state updates. We instantiate this paradigm in ViLaVT, a novel LVLM equipped with a dynamic vision encoder explicitly designed for such interactive visual reasoning, and trained it with a two-stage curriculum combining supervised fine-tuning and reinforcement learning to promote effective reasoning behaviors. Extensive experiments across eight benchmarks demonstrate that ViLaVT achieves strong and consistent improvements, with particularly pronounced gains on complex multi-image and video-based spatial reasoning tasks.

---