EngramBench is a new benchmark designed to evaluate skill evolution in autonomous agents by focusing on genuine capability abstraction rather than solution copying. It includes 30 learning tasks and 13 unseen transfer tasks that require agents to manage complex, multi-hour development cycles with LLM‑simulated users. The study shows that while static skill banks cannot eliminate the need for precise code implementation, they effectively reduce redundant context and cut overall coding time by more than 55%.
By Zhixuan Tan, Pengjie Gu, Zhao Li, Yihan Hu, Xu He, Dong Li, Jianye Hao
The paper introduces a Time-Aware Bag-of-Receptive-Fields (BORF) for classifying irregular time series, extending the original BORF to handle non-uniform sampling, missing data, and variable lengths. It adds a time-weighted normalization that weights observations by their time deltas, enabling pattern extraction that reflects the true temporal distribution. The method maintains linear time complexity and is evaluated against state‑of‑the‑art irregular time‑series classifiers, achieving competitive performance while providing human‑interpretable explanations.
By Francesco Spinnato
The paper introduces FDCU, a dual‑constrained subspace projection framework designed to improve machine unlearning for large language models. FDCU limits parameter updates with a dual‑masking rule that preserves general knowledge via Fisher Information while preventing the activation of spurious suppressors through the Principle of Minimal Functional Intervention. Experiments show that FDCU achieves state‑of‑the‑art robustness against retraining attacks while maintaining near‑lossless general utility, thereby ensuring durable safety for LLMs.
By Jiaqing Li, Shide Zhou, Zhibo Zhang, Yuxi Li, Tianlong Yu, Kailong Wang
WinoTS introduces a wavelet‑based self‑distillation framework for time‑series models that uses time‑frequency augmentations to create multi‑scale structural views, avoiding distortion of signal dynamics. The method outperforms state‑of‑the‑art baselines in long‑term forecasting, cross‑domain zero‑shot transfer, and unsupervised anomaly detection, and linear probing on frozen representations often beats fully supervised training from scratch. Ablation studies show WinoTS is architecture‑agnostic and demonstrates that time‑frequency transformations offer a principled alternative to vision‑style spatial augmentations.
By Noam Major, Kathy Razmadze, Yoli Shavit
The paper introduces Mosaic, a multi-image visual harness that lets large language‑vision models (MLLMs) construct visual intermediates using ten composable image operations. It evaluates five re‑representation settings on existing multi‑image benchmarks and a new grounding‑focused benchmark, MosaicBench, finding that visual re‑representation benefits tasks requiring precise visual evidence more than those dominated by high‑level semantics. MosaicAgent‑8B is trained via reinforcement learning to compose these operations without demonstration trajectories, demonstrating diverse problem‑solving patterns.
By Gengyuan Zhang, Xiao Han, Xinyu Xie, Tong Liu, Volker Tresp
CAST introduces a reinforcement‑learning fine‑tuning framework for diffusion models that addresses three key limitations: it automatically selects the denoising window based on each model’s trajectory, decomposes prompts into verifiable semantic atoms via Causal Scene Graphs, and applies atom‑level rewards spatially weighted in the policy objective. The method is applied to FLUX.2‑dev and Qwen‑Image‑2512, yielding up to 3.07× improvement on the hardest GenEval 2 prompts compared with Flow‑GRPO while also enhancing overall generation quality.
By Shu Yu, Chaochao Lu
GroundingPI is a 4‑billion‑parameter grounding foundation model that generates points and boxes as quantized coordinates in a shared vocabulary. It is trained with multimodal and spatial pretraining, supervised fine‑tuning, and reinforcement learning, achieving a new state‑of‑the‑art average of 73.68% across 34 grounding benchmarks. As a visual backbone, GroundingPI improves performance in robotic manipulation and autonomous driving, outperforming larger models and mainstream backbones in several out‑of‑distribution settings.
By Qize Yu, Lianrui Fan, Boyu Chen, Jiaqi Liang, Xini Ding, Yue Chen, Zetian Song, Yuran Wang, Yi Zou, Kaixuan Wang, Tianxing Chen, Wenxuan Song, Bohan Zhou, Mingleyang Li, Siqiao Huang, Yuqi Ye, Caigao Jiang, Wei Wei, Ruihai Wu, Hang Zhang, Yixiao Ge, Shuchang Zhou, Shilong Liu, Xianming Liu, Ping Luo, Shiyu Huang
D‑Scope is a framework that links the interpretation of sparse autoencoder (SAE) features in diffusion transformers (DiTs) to controllable image generation. It aggregates SigLIP‑2 embeddings of highly activating image patches into visual centroids, matches target text descriptions against these centroids, and retrieves individual features without per‑feature text annotations. The method provides visual evidence for each selection and uses spatially masked interventions to test decoder directions under fixed generation conditions, evaluated across 150 SAEs and a benchmark of 100 target concepts.
By Xinyue Xu, Jiahao Zhang, Lijie Hu, Peter Hase, Hao Wang
The paper introduces FailBank, a four‑stage self‑evolving framework that transforms runtime feedback from safety shields into lasting policy improvements for vision‑language‑action (VLA) models. By using a counterfactual correction teacher, outcome‑aware admission, and guarded LoRA updates, FailBank converts useful shield proposals into corrective targets while preserving successful actions as anchors. Experiments on the VLA‑Arena benchmark show that FailBank boosts task success rates by up to 8.5 percentage points and reduces cumulative policy cost by up to 35.6%, outperforming both base policies and traditional runtime shielding.
By Mingyue Cui, Zheyuan Liu, Yihan Zhu, Zheyuan Zhang, Meng Jiang
TRACE is a lightweight learned selector that ranks completed search trajectories by aggregating cross‑rollout evidence, preserving individual query and evidence occurrences while propagating information across shared content or document identity. Trained with answer‑level supervision over frozen text embeddings, TRACE selects an existing answer without additional search or autoregressive aggregation, and a single selector generalizes across rollout policies and agent backbones. Across six WebQA policies, six long‑horizon dataset‑backbone combinations, and multiple WebQA benchmarks, TRACE outperforms majority voting and generative aggregators, achieving higher accuracy and at least tenfold higher processing throughput.
By Qisheng Zhou, Zhen Xiong, Qiaoyu Tan
LEAP is a framework for long audio‑video question answering that avoids encoding entire recordings by dividing them into fixed‑duration blocks. It performs a lightweight localization pass on each block to score short candidate windows, then pools the highest‑ranked windows for a single bounded answer pass, keeping the answer input and peak context independent of recording length. The method trains both a localization LoRA and an answer LoRA, supports causal streaming queries, and achieves significant performance gains over baseline models on multiple AVQA benchmarks.
By Juyi Lin, Zhiqiang Lao, Jiali Cui, Lin Zhao, Pu Zhao, Dichang Zhang, Arman Akbari, Yu Qi, Xinru Jiang, Yanzhi Wang, Heather Yu, Liang Peng
Mid‑Harness proposes a test‑time compute strategy that samples and verifies candidate actions before execution, keeping the underlying generator and harness unchanged. Experiments show that with a strong verifier, sampling more actions significantly boosts success rates—e.g., a GPT‑5.6 verifier raises Pass@1 from 50.00 % to 68.03 % on TerminalBench‑Lite using eight samples. The approach also improves performance across various models, benchmarks, and harnesses, demonstrating that action scaling is a promising target for enhancing terminal agent reliability.
By Minki Kang, Ryo Hachiuma, Shaokun Zhang, Subhashree Radhakrishnan, Yonggan Fu, Jindong Jiang, Mingjie Liu, Ehsan Hosseini-Asl, Yi Dong, Yu-Chiang Frank Wang, Byung-Kwan Lee
PhantomEnvironments is a framework that trains large language model agents in synthetic, rule‑generated fictional worlds. By creating multi‑turn reinforcement learning environments where agents search templated articles to answer multi‑hop questions, the approach eliminates the need for costly human data or hallucinated LLM‑generated settings. Agents trained in these zero‑cost, purely rule‑based worlds transfer effectively to real‑world multi‑hop search benchmarks, often surpassing models trained on real data, and demonstrate scalable search behavior that grows linearly with question difficulty.
By Anmol Kabra, Swathi Saravana Selvam, Albert Gong, Chao Wan, Christian Belardi, Dongyoung Go, Katie Z. Luo, Kilian Q. Weinberger
arXiv:2609.40322v1 Announce Type: cross
Abstract: Material generation should produce not only an appearance, but also the rules that construct it. We introduce MatLoom, a compact, layer-oriented lang...
By Anson Y. Lam, Shuqing Li, Michael R. Lyu
arXiv:2609.40360v1 Announce Type: cross
Abstract: Reinforcement learning with verifiable rewards (RLVR) has improved the reasoning capabilities of large language models (LLMs), yet their predictions...
By Junshu Pan, Zhizhang Fu, Shulin Huang, Yiran Ding, Zifan Cheng, Wenqi Shao, Qiaosheng Zhang, Yue Zhang
arXiv:2609.35561v2 Announce Type: replace
Abstract: Recursive self-improvement (RSI) seeks to enable AI systems to participate in improving their own capabilities. A concrete pathway is autonomous mo...
By Yaxin Du, Xiyuan Yang, Zhifan Zhou, Yujie Ge, Cheng Wang, Jiajun Wang, Sijie Chen, Zehui Liu, Yuxin Zhang, Weicheng Gu, Julian Zhang, Zixing Lei, Siheng Chen
arXiv:2608.24764v2 Announce Type: replace
Abstract: As language-model agents become more capable of iterative search, corpus access is shifting from retrieval toward interaction. Agents can explore t...
By Hongyu Guo, Zhiyu Zheng, Minghao Hu, Shuai Lei, Zhunchen Luo, Chao Ma, Yunbo Cao, Zhao Cao
arXiv:2609.00768v2 Announce Type: replace
Abstract: Self-play supports the self-evolution of language models, but solver performance can plateau or decline across rounds without guidance. Existing un...
By Xincheng Wei, Yifan Ding, Fucheng Xiong, Yoshua Li, Dongsheng Ma, Rongxiang Weng, Xunliang Cai, Wenjian Ding, Yao Zhang
arXiv:2609.33665v2 Announce Type: replace
Abstract: Automatically generated environments provide a scalable source of interaction data for training general agents. However, existing approaches mainly...
By Xiao-Wen Yang, Weiyi Xu, Wen Da, Hang Xu, Canwei Li, Hong-Jie You, Pusen Dong, Yucheng Zeng, Zhaokai Luo, Yu-Feng Li, Yao Hu, Mu Chuan
arXiv:2609.37054v2 Announce Type: replace
Abstract: Ensuring the safety of reasoning large language models (LLMs) across languages is essential for their reliable deployment. However, when exposed to...
By Xianhui Zhang, Jian Yu, Chengyu Xie, Chenhang Cui, Shuyi Miao, Pengyang Shao, Yu Zheng, Fei Shen, Tat-Seng Chua