QEMem, a framework that reformulates memory utilization as the construction of query-specific evidence sets, shifting the focus from individual retrieval relevance to collective evidential support, achieves improved evidence coverage and strong performance on questions requiring cross-session evidence integration.
Shi-Xuan Li, Wei Yang, Pei-Yu Zhang et al.· 0 citations
This work proposes minimal-edit corrective on-policy distillation, a method that preserves on-policy supervision while avoiding full-response imitation, and improves perception-oriented performance, particularly low-level geometric perception, and transfers to external mathematics and perception benchmarks.
Wen Ye, Ai-Lin Deng, Zi-Rui Wang et al.· 0 citations
Looped language models reuse physical Transformer blocks across logical depth, but established architectures entangle recurrence placement, input injection, and recurrent state organization. We isolate these axes in models trained from scratch at matched parameter scale, logical depth, and token budgets, and then test...
Benhao Huang, Chu-Fan Shi, Jun-Lin Chen et al.· 4 citations
Self-evolving LLM agents promise to improve autonomously through continual interaction and learning, reducing their dependence on manually curated supervision. Realizing this promise requires not only updating the agent, but also evolving its training experience as its capabilities change. However, most existing pipeli...
Unified multimodal models (UMMs) integrate understanding and generation, yet their generative behavior is rarely governed by what they understand about the task. We formalize visual abstention: when a requested visual transformation is impossible under the task's rules, the model should recognize that no valid solution...
Chu-Fan Shi, Cheng Yang, Tian-Nuo Yang et al.· 0 citations
This work improves the two components of training that shape these fixed points: the depth prior and input injection, and learns the prior from prediction feedback, with an entropy term that keeps it broad.
This work proposes recursive cross-capability self-improvement (RSI), a training loop in which the text and visual abilities of a UMM supply training data for one another, in which the text and visual abilities of a UMM supply training data for one another.
Hui-Juan Wang, Chu-Fan Shi, Cheng Yang et al.· 1 citation· ⚡1
Multimodal foundation models edit natural photographs at production quality, yet the same models struggle with structured visual content such as infographics. Unlike photographs, infographics encode information through logical relations; editing one element often requires surrounding elements to be adapted. We refer to...
Cheng Yang, Chu-Fan Shi, Hui-Juan Wang et al.· Proceedings of the 2026 Conf...· 0 citations
K2-V2 is introduced, a 360-open LLM built from scratch as a superior base for reasoning adaptation, in addition to functions such as conversation and knowledge retrieval from general LLMs, to empower the community with a capable, reasoning-centric foundation.
K2 Team Zhengzhong Liu, Liping Tang, Ling-Hao Jin et al.· 0 citations
We use cookies to run the site and, with your consent, for analytics and to show ads.
See our Cookie Policy.