…Notably, 17 out of 20 tested models perform worse than a random-guess baseline (15.2%) on Gem Set, while the estimated human performance is ~81.1%, showing substantial room for improvement…
…On several benchmarks, FEST outperforms baselines with magnitudes less SFT data, even matching their performance with full dataset. View arXiv page View PDF GitHub 0 Add to collection Community This is a…
…Experiments show that current vision backbones encode semantic structure but struggle with cross-category correspondence and object-part position, while LVLMs perform better at text-prompted localization than visual-reference matching. SOCO…
…Rather than eliminating test-time world evolution altogether or rolling out the full future in pixel space, DAWN performs a short explicit latent rollout that is sufficient to support long-horizon trajectory…
…training that uses orthogonal equivalence transformations to maintain singular values during weight updates, offering stable performance comparable to standard optimizers. AI-generated summary We introduce Pion, a spectrum-preserving optimizer for large…
…Jiaqi Tang , , , , , , , , Abstract Robust-U1 enhances multimodal large language models' robustness against visual corruptions through self-recovery capabilities that improve both visual quality and reasoning performance. Generated by Qwen/Qwen2.5-Coder…
…Yanggan Gu , Shuo Cai , , Wenjun Wang , Yuanyi Wang , , , , , Abstract Feature drift analysis in model merging leads to FeatCal, a calibration method that reduces performance gaps through layer-wise weight updates without gradient…
…Our framework serves as a practical and reliable evaluation tool for diagnosing content consistency and benchmarking model performance in real-world I2I systems. View arXiv page View PDF Project page GitHub 18…
…Experiments demonstrate that Light-WAM maintains strong performance on LIBERO and achieves usable multi-task performance on RoboTwin 2.0, while using only 0.44B trainable parameters. It also achieves 72.03ms…
…With long-context models, many-shot ICL can use dozens to hundreds of examples and achieve performance comparable to fine-tuning, yet current understanding of its scaling behavior is largely derived from…