RegCL is a non‑replay continual learning framework that adapts the Segment Anything Model (SAM) for visual grounding across evolving multi‑sensorial media domains. It consolidates domain‑specific segmentation knowledge into a single lightweight SAM adapter by incrementally merging LoRA‑style AugModules and preserving compact historical feature statistics. Experiments on five heterogeneous datasets demonstrate that RegCL retains performance while adapting to new domains, outperforming other non‑replay continual learning and merging baselines.
By Yuan-Chen Shu, Zhiwei Lin, Xiaoyu Zhou, Yongtao Wang
The paper introduces a new protocol, Mistake Detection Video Question Answering (MD‑VQA), to evaluate whether models can determine if a step in a video follows its description, covering both seen and unseen actions. It proposes a post‑training approach for video‑language models that uses a reward function to highlight discrepancies between instructions and video content. Experiments show this method surpasses zero‑shot, fine‑tuned, and other post‑training baselines, especially on unseen procedures, improving performance by up to 11.6% on EP‑VQA.
By Federico Spurio, Olga Zatsarynna, Lars Doorenbos, Emad Bahrami, Gianpiero Francesca, Juergen Gall
arXiv:2601.13247v2 Announce Type: replace-cross
Abstract: Current Large Language Models (LLMs) exhibit a critical modal disconnect: they possess vast semantic knowledge but lack the procedural ground...
By Baochang Ren, Yunzhi Yao, Rui Sun, Shuofei Qiao, Ningyu Zhang, Huajun Chen
arXiv:2606.21359v2 Announce Type: replace
Abstract: Large language models (LLMs) are increasingly used to communicate and explain scientific concepts, yet their tendency to hallucinate poses signific...
By Raia Abu Ahmad, Nikolas Rauscher, Ekaterina Borisova, Fabio Barth, Georg Rehm, Sebastian M\"oller
The paper investigates cultural biases in large language models (LLMs) by introducing the Culture-Related Open Questions (CROQ) dataset, which contains 24‑language questions about generic culture. Experiments reveal that LLMs disproportionately favor Japan in their responses, especially when prompted in high‑resource languages, while low‑resource languages tend to highlight countries where the language is official. The study also finds that these biases emerge after supervised fine‑tuning rather than during pre‑training.
By Joseba Fernandez de Landa, Carla Perez-Almendros, Jose Camacho-Collados
Supervised fine-tuning (SFT) is the de facto standard for adapting large language models (LLMs) to target domains, but it often degrades the model's general capabilities, a phenomenon known as catastr...
Detecting hallucinations in Large Vision-Language Models (LVLMs) requires both accurate span localization and well-calibrated confidence scores. Fine-tuned generative VLMs excel at identifying halluci...
The paper introduces SWIFT, a Swin V2‑based model pretrained on 10,444 3D CT volumes and fine‑tuned for rectal cancer segmentation on T2‑weighted MRI. Four configurations—full fine‑tuning (SWIFT), decoder compression (SWIFTe), low‑rank adaptation (SWIFTe‑LoRA), and a LoRA‑decoder ensemble (SWIFTe‑LDE4)—were evaluated on 247 cases, showing that SWIFTe reduces parameters by 70.1% while improving tumor detection and radiomic agreement. The study also demonstrates a trade‑off between detection and boundary agreement, and highlights that SWIFTe‑LDE4 achieves the lowest calibration errors after temperature scaling.
By Aneesh Rangnekar, Jorge Tapias Gomez, Joseph O Deasy, Harini Veeraraghavan
The paper proposes a scalable, automated method for auditing candidate‑job matching systems for demographic bias. It employs large‑language‑model agents to generate neutral resumes, injects controlled demographic variations, ranks candidates with a fine‑tuned embedding model, and evaluates nine fairness metrics across counterfactual, group‑fairness, and merit‑aware families, producing a composite risk report. Experiments on a small corpus show that single‑score audits miss nuanced issues, underscoring the need for multi‑metric evaluation and LLM‑generated audits as a low‑cost complement to human reviews.
By Sai Yashwant, Shruti Bansal, Anurag Dubey, Samaroha Chatterjee, Satyam Kumar, Shreyash Gupta, Gantala Thulsiram
The paper introduces MCCE, a hybrid framework that combines a frozen closed‑source large language model (LLM) with a lightweight, trainable model for multi‑objective discrete optimization. By maintaining a trajectory memory and refining the small model through reinforcement learning, the two models jointly enhance global exploration and learning. Experiments on drug‑design benchmarks demonstrate that MCCE achieves state‑of‑the‑art Pareto front quality, outperforming existing baselines.
By Nian Ran, Zhongzheng Li, Yue Wang, Qingsong Ran, Xiaoyuan Zhang, Shikun Feng, Richard Allmendinger, Xiaoguang Zhao
The paper introduces Thomson, a frontier AI model developed through continual learning on open-weight models, aiming to democratize access to high-performance AI. It argues that institutions with limited resources can achieve frontier-level performance by applying a modern mid- & post-training stack, preserving model plasticity and stability while minimizing high-impact interventions. Thomson demonstrates competitive performance across agentic tasks, safety, legal, tax, multilingualism, and large-scale deep research, exhibiting a distinctive π-shaped improvement pattern and effectively mitigating the forgetting problem seen in narrow domain adaptation.
By Shengzhuang Chen, Jerrod Parker, Yejin Bang, Andrew M. Bean, Nabeel Seedat, Stefan Winzeck, Daniil Glazko, Jannik Zgraggen, Fangyi Yu, Scott Arnott, Dietrich Trautmann, Luca Ciuffreda, Guglielmo Bonifazi, Davide Romano, Bradley Bell, Kirsty Fielding, Daniele Giofr\`e, Tom Zielund, Ipshita Chatterjee, Sneha Murthy Ghantasala, Manpreet Nanreh, John Scoville, Maciej Sakowicz, Wassim Seifeddine, Lukas Thede, Jonathan Richard Schwarz
Naive Prompt Optimization (NPO) is a lightweight, single‑lineage method that iteratively refines prompts using a teacher model’s rollout feedback. It matches or surpasses the performance of more complex optimizers like GEPA while requiring fewer rollouts, and its advantage grows with stronger teacher models. In interactive games, NPO remains competitive, and prompts optimized by NPO transfer well to other student models within the same family.
By Yuan Chang, Xiaoqi Chen
The paper investigates how transformers acquire deep semantic dependencies, proposing a mechanistic framework that frames learning as a competition between surface statistics and deep semantics. It identifies a "Gradient Starvation" effect that suppresses error signals for sparse semantic dependencies early in training, delaying structural reasoning until a sudden phase transition. The study also explains the success of Chain-of-Thought strategies and introduces a topology‑aligned contrastive objective that improves variable binding performance by more than twice the gain of standard fine‑tuning.
By Jiangrui Zhao, Xiaoting Du
The paper introduces a low‑rank auditing method called LoRA as Oracle, which fits a small adapter to a hypothesis and analyzes the geometry, energy, and alignment of the resulting update relative to frozen weights. This approach directly measures what a model has internalized, independent of its output behavior, enabling detection of backdoors that behavioral audits miss. By identifying and erasing malicious internalizations within the same low‑rank subspace, the method consistently detects target classes across multiple datasets and architectures while preserving clean accuracy and operating at far lower parameter and memory cost than full‑model baselines.
By Marco Arazzi, Antonino Nocera
The paper proposes a method called Modular Expert Merging for Biomedical Retrieval, which combines independently trained domain‑specialized experts instead of large mixed‑domain training. Experiments across four decoder‑only LLM families (0.6B‑7B) and twelve retrieval tasks from MTEB show that merging experts consistently outperforms mixed‑domain training. The authors also introduce a Synthesize‑Train‑Merge (STM) framework that generates hard negatives with a top‑tier LLM, fine‑tunes experts via LoRA, and merges them, achieving strong biomedical retrieval performance while retaining competitive general‑domain results.
By Sameh Khattab, Jean-Philippe Corbeil, Osman Alperen \c{C}inar-Kora\c{s}, Amin Dada, Julian Friedrich, Jiawei He, Douglas Teodoro, Jens Kleesiek
The paper investigates how the order of candidate documents in a prompt affects the decisions made by large‑language‑model (LLM) scorers, even when their ranking quality is similar. It shows that five scorers with only a 0.010 nDCG@10 difference can produce retained‑set overlaps as low as 0.66–0.84, and that existing rerankers still exhibit significant order dependence. The authors propose Order‑Consistency SFT (OC‑SFT), a training method that reduces this dependence, maintaining ranking quality while improving decision stability across multiple tasks.
By Markus Frohmann, Mahdiyar Alavi, Elizabeth Lingg, Navid Rekabsaz
Refine-POI introduces a reinforcement fine-tuned framework for next point-of-interest recommendation that tackles two key issues: topology-blind indexing of semantic IDs and the limitation of supervised fine-tuning to top‑1 predictions. It uses a hierarchical self‑organizing map to generate topology‑aware semantic IDs and a policy‑gradient approach to produce top‑k recommendation lists. Experiments on three real‑world datasets show that Refine‑POI outperforms state‑of‑the‑art baselines, combining LLM reasoning with accurate, explainable recommendations.
By Peibo Li, Shuang Ao, Hao Xue, Yang Song, Maarten de Rijke, Johan Barth\'elemy, Tomasz Bednarz, Flora D. Salim
The paper introduces GRAPHSU, a graph‑guided selective unlearning method for language models that expands deletion beyond explicitly identified forget seeds. By constructing a weighted support‑route graph and propagating deletion pressure, GRAPHSU applies graded forgetting to high‑risk neighboring examples. Experiments on the TOFU and PISTOL benchmarks with GPT‑2 Medium and Llama‑3.2‑3B‑Instruct show that GRAPHSU achieves the lowest utility‑feasible soft leakage, reducing leakage by up to 49.5 percentage points compared to a seed‑only baseline.
By Waqas Khan, Tabinda Sarwar, Jingyue Cong, Xun Yi, Estrid He
MemToC is a controlled benchmark that tests how large language models resolve conflicts between their internal memory and tool outputs. It contains 6,504 episodes built from 542 factual questions, each paired with a model‑generated closed‑book answer and a tool return whose correctness is known, creating four distinct source‑correctness scenarios. Across five 7‑9B open‑weight models, tool responses overwhelmingly dominate closed‑book answers, and only a minority of instruction‑tuned models correctly retain a verified answer when the tool is wrong, while most follow a correct tool or repeat a wrong tool.
By Arseniy Varlamov, Rishat Zinnatullin, Elisei Rykov, Alexander Panchenko, Ilseyar Alimova
The survey reviews how Large Language Models (LLMs) are being used in High‑Performance Computing (HPC) programming, covering code generation, parallelization, frameworks, evaluation, and broader challenges. It finds that general‑purpose LLMs perform adequately on serial and OpenMP‑style tasks but struggle with distributed MPI workloads, while domain‑specialized models achieve higher accuracy yet are limited in scope and evaluation. The authors argue that LLMs will not replace HPC experts soon but can act as powerful collaborators, provided richer datasets, integration with performance tools, rigorous evaluation, and governance are developed.
By Strahinja Ljaljevic, Josep Jorba, Sergio Iserte