The paper critiques current encoded‑prompt safety benchmarks that focus only on harmful requests, showing that such tests can misrepresent a model’s safety. By evaluating the benign arm under the same encoding, the authors reveal a substantial drop in the harm gap—sometimes to zero—indicating that the encoding masks true safety deficiencies. Across multiple large models and training pipelines, they document that the encoding can either hide or falsely inflate safety metrics, and they identify twelve specific instrument defects that contribute to these misleading results.
By Haoyu Zhang, Haowen Xu, Xiao Luo, Hanwen Liu, Yang Chen, Zijian Xiao, Yi Feng, Xiangchen Guan, Mohammad Zandsalimy, Shanu Sushmita
This study benchmarks transformer models for Bangla medical named entity recognition (NER), comparing BanglaBERT, multilingual BERT (mBERT), XLM‑RoBERTa, and GPT‑4o mini under zero‑shot and few‑shot prompting. Across a full test set of 3,179 samples, fine‑tuned XLM‑RoBERTa achieves a new state‑of‑the‑art F1‑score of 0.5959, while BanglaBERT lags with 0.4937, suggesting that domain diversity outweighs language specificity. The analysis shows high performance on Medicine and Specialist entities (F1 > 0.83) but lower accuracy on Symptoms (F1 0.4367), and demonstrates that fine‑tuned transformers outperform prompt‑only approaches by a factor of 3.76.
By Rakib Abdullah, Md. Maruful Islam Maruf
The paper presents a graph-based framework for large-scale railway network management, combining a hierarchical Bayesian model with a Gaussian Process on a graph kernel to infer spatially correlated maintenance environments from Swiss Federal Railways data. It introduces a topology-aware Multi-Agent Reinforcement Learning system that uses graph neural networks and Transformers to optimize network-level policies. The approach demonstrates scalability via zero-shot transfer learning, enabling agents trained on small network segments to perform effectively on unseen large networks, outperforming heuristics and standard MARL baselines while reducing training time.
By Giacomo Arcieri, Gregory Duth\'e, Christophe Muller, Konstantinos G. Papakonstantinou, Daniel Straub, Eleni Chatzi
The paper presents an inference-only pipeline that extends the frozen NER model MahaNER‑BERT to document‑level prediction using overlapping sliding windows, eliminating the need for retraining or architectural changes. The approach is evaluated on six document‑level corpora derived from the MahaNER test set, employing two repetition strategies (Normal Repeat and Random Repeat) at three length levels and various window configurations. Results show the model maintains a macro F1‑score of up to 0.8902 with minimal variation, outperforming non‑windowed methods by avoiding boundary‑fragmentation errors and achieving more stable document‑level performance.
By Hariom Ingle, Ronit Ghode, Ishwari Gondkar, Jidnyasa Harad, Ravindra Murumkar, Raviraj Joshi
EAGER is a reinforcement learning framework for generative event extraction that uses fine‑grained verifiable rewards and Schema‑Contrastive Advantage Estimation to address advantage collapse in sparse binary reward settings. The reward design focuses on structural validity, extraction accuracy, groundedness, coverage, over‑generation, and span precision. Across seven benchmark datasets, EAGER outperforms prompting, supervised fine‑tuning, and previous reinforcement learning baselines, achieving a substantial improvement over the strongest prior method.
By Omar Adjali, Siting Liang, Omair Shahzad Bhatti, Daniel Sonntag
The study re‑examines a reported advantage of a routed ternary (1.58‑bit) language model over a full‑precision transformer at 60K parameters. By running controlled experiments with multiple seeds and a fixed training recipe, the authors find that the apparent benefit largely stems from the choice of baseline model shape rather than the ternary architecture itself. While the routed model does outperform other shapes at a larger 130M‑byte budget, its advantage diminishes when a plain gated diagonal‑SSM block is used, and the ternary penalty varies with architecture and quantization details.
By Gautam Veldanda
The paper investigates how the order of generating explanations—whether a rationale is produced before or after the answer—affects vision‑language reasoning. By conducting controlled experiments on knowledge‑intensive QA, visual entailment, and compositional grounding tasks, the authors show that larger models are required for reliable rationale‑first generation, while answer‑first generation is less susceptible to format errors. The study concludes that explanation ordering, model scale, pre‑training knowledge, fine‑tuning, and task structure jointly influence prediction accuracy and reasoning faithfulness.
By Siting Liang, Luca Rippe, Omar Adjali, Daniel Sonntag
The study evaluates Arabic–Russian machine translation by comparing seven fine‑tuned neural machine translation (NMT) models with four few‑shot large language models (LLMs) on a new 15.47 million‑pair corpus split into 20k/5k/5k. Fine‑tuned NLLB‑1.3B achieves the best performance (BLEU 16.3, COMET 0.738), while the best few‑shot LLM, Aya‑Expanse 8B, scores only BLEU 1.7 on 500 sentences. Error analysis shows that low lexical overlap between Arabic and Russian is the main source of failures, and statistical tests confirm significant performance gaps between most models.
By Mullosharaf K. Arabov
The paper introduces task-informed parameter-efficient fine-tuning methods for low-resource speech recognition by applying Fisher-Whitened Cross-Covariance Analysis (FCCA) to Whisper and Qwen3-ASR. Two extensions—Asymmetric-Coupled FCCA (AC‑FCCA) and Adaptive‑Rank FCCA (AR‑FCCA)—are proposed to exploit cross‑layer sharing and adapt rank allocation within a fixed parameter budget. Experiments on multilingual datasets show that standard FCCA matches or surpasses LoRA, while AR‑FCCA consistently improves performance across models without increasing trainable parameters.
By Asmee Mishra, Mengjie Qian, Brechtje Post, Kate Knill
The paper introduces a perception interface that separates vision from language in vision‑language models. A frozen perception stack detects objects, a deterministic semantic serializer converts the perceived state into text, and a standard text‑only large language model (LLM) answers questions. Experiments on a campus‑robot benchmark show that this serialized interface outperforms a zero‑shot VLM of the same language‑model size, especially as the language model shrinks, and that the advantage persists under paraphrase and different supervision regimes.
By Cong Xu, Ravi Sankar
The paper presents a 3D foundation model for light sheet fluorescence microscopy (LSM) that is pretrained on a large curated set of 3D images from various organisms, stains, and imaging protocols. By jointly optimizing for masked reconstruction and image‑text alignment, the model learns transferable volumetric representations that dramatically reduce the need for annotated data. The pretrained backbone enables efficient few‑shot adaptation to downstream tasks such as segmentation, classification, and deblurring, consistently outperforming baselines according to standard metrics and expert evaluation.
By Adina Scheinfeld, Haotan Zhang, Shang Mu, Rudolf L. M. van Herten, Lucas Stoffl, Ali Erturk, Zhuhao Wu, Johannes C. Paetzold
The paper introduces Smol‑VL‑BLV, a compact vision‑language model designed for blind and low‑vision users. It employs a 500M decoder transformer with teacher‑student distillation and Group Relative Policy Optimization to add spatial detail, directional cues, and hazard detection to post‑training. After a lightweight finetuning step, the model achieves significant gains on spatial, social, OCR, and VQA benchmarks while remaining under 450 MB and running entirely offline on a mid‑range Android phone.
By Rishabh Choudhary, Shreyansh Raj, Umesh Goyal, Shubh Kashyap, Shrestha Kumar, Sushovan Jena, Komal Kumar, Hisham Cholakkal, Aditya Nigam
The paper addresses look‑ahead bias in large language models (LLMs) used for financial prediction, which arises because LLMs are trained on long time‑series data. It proposes a low‑cost solution that adjusts the logits of a base model at inference time using two smaller, specialized models—one fine‑tuned to forget certain information and another to retain it. Experiments show that this method removes both verbatim and semantic knowledge, corrects biases, and outperforms previous approaches.
By Humzah Merchant, Bradford Levy
CounterRoute is an online reinforcement‑learning framework that jointly learns how to route a language model’s reasoning between a ‘think’ and a ‘direct answer’ mode, using a single shared policy derived from a dual‑mode checkpoint. It employs counterfactual rollouts to credit routing decisions and a curriculum that starts with forced dual‑mode rollouts before shifting to self‑routed updates, achieving better accuracy‑efficiency trade‑offs across nine benchmarks. The method reduces generated tokens by up to 51% on Qwen3‑8B while improving macro‑average accuracy, and its routing strategy generalizes to unseen coding, science, and commonsense tasks.
By Ruochen Jiao, Besnik Fetahu, Zhenyu Shi, Priyanka Nigam
The paper "Learning to Ideate for Scientific Impact" explores using delayed signals of scientific uptake—specifically citation-normalized impact—as feedback to steer large language models toward generating high‑impact research ideas. The authors build a dataset of over 100,000 computer science papers, train a reward model to predict citation impact from goal‑idea pairs, and align an idea generator via supervised fine‑tuning and reinforcement learning. Evaluation with a reference‑grounded protocol shows that the RL‑tuned model consistently produces ideas with higher estimated impact than baseline models.
By Shubham Kale, Aniketh Garikaparthi, Manasi Patwardhan
TraceGuard is an adaptive, rank‑based filtering method designed to detect and remove poisoned image‑text pairs from multimodal training corpora. It evaluates six corpus‑level features—such as cross‑modal neighborhoods, recurring text, and changes after text‑span erasure—to identify suspicious examples without training the victim model. In experiments across 19 attack scenarios, TraceGuard successfully removes an average of 98.4% of poisoned data while discarding only 5.4% of clean data, reducing residual attack impact to at most 1% in most configurations.
By Haoyang Li, Yaxin Xiao, Linyan Dai, Jiawen Fu, Zi Liang, Jason Xue, Qingqing Ye, Haibo Hu
GridSFM is a 15‑million‑parameter physics‑inspired graph neural network that serves as a foundation model for solving AC Optimal Power Flow (AC‑OPF) across diverse grid topologies. Pretrained on 54 topologies ranging from 500 to 4,000 buses, it achieves a 2.45 % zero‑shot generation‑cost error on a held‑out 10,000‑bus case and adapts to unseen grids with only 100 solved instances using a physics‑informed fine‑tuning scheme based on Newton’s method. The authors address the disconnected feasible set of AC‑OPF by lifting and relaxing constraints with logarithmically penalized slacks, proving the resulting elastic feasible set is contractible and that solutions can be projected back onto the original feasible set.
By Luke Bhan, Weiwei Yang, Margaret Capetz, Baosen Zhang
The paper introduces a search‑aware reinforcement learning framework for multi‑component query understanding in Roblox game search. It first uses teacher‑student supervised fine‑tuning to create a schema‑compliant policy, then applies reinforcement learning that optimizes each query‑understanding component with component‑specific rewards derived from live search engine interactions. Experiments show that this approach improves per‑component utility and overall search quality, raising NDCG@20 by 8.9 points over the supervised baseline and 3.5 points over a single end‑to‑end reward strategy.
By Nayoung Choi, Shengjian Chen, Xiaokai Wei, Wenzheng Zhang, Daiyao Yi, Rachit Pareek, Vincent Su, Michelle Gong, Jinho D. Choi
HarnessPAI is a model‑ and embodiment‑agnostic framework that treats code as an executable, evolvable interface for Physical AI. It separates short‑term open‑loop program execution from long‑term closed‑loop evolution, using feedback to refine programs and distill reusable skills. Across various robots, HarnessPAI outperforms pure action models and code‑as‑policy baselines, achieving significant gains on tasks like LIBERO‑PRO and RoboCasa without retraining the underlying model.
By Xin Wang, Wenhao Wu, Menghao Zhang, Zhi Wang, Kun Shao, Jian Luan, Yang Li, Qing Li, Shangding Gu, Huichi Zhou, Shuqing Shi, Fei Ni, Shuo Lu, Weicheng Meng, Kang Li, Jin Wu, Kang Zhao, Shangmin Guo, Gen Li, Yongqiang Tang, Zhizhong Zhang, Yuan Xie, Heng Qu
The paper presents a graph-based framework for large-scale railway network management that combines a hierarchical Bayesian model with a Gaussian Process on a graph kernel to model spatially correlated maintenance environments, and a topology-aware Multi-Agent Reinforcement Learning system using graph neural networks and Transformers to optimize network-level policies. It demonstrates scalability by training agents on small network segments and deploying them zero-shot on larger, unseen networks, achieving superior performance over heuristics and standard MARL baselines while reducing training time. The approach addresses the computational challenges of centralized methods and the coordination gaps of decentralized methods in complex, long-horizon infrastructure asset management.