Large language model (LLM) agents rely on memory systems to leverage long-term interactions, yet existing systems typically retrieve memories based on their relevance to the current query and use them as contextual evidence. However, relevance alone does not characterize how a memory contributes to task solving: the sa...
Mingyu Zhang, Zi-Xu Li, Zhi-Heng Fu et al.· 0 citations
Rubric-based reinforcement learning uses fine-grained grader feedback to train large language models on open-ended tasks, but grader randomness can alter GRPO updates in ways that score discrepancies alone cannot characterize. Because GRPO jointly normalizes rewards within each response group and combines the resulting...
Jia-Le Huang, Zi-Xu Li, Zhi-Wei Chen et al.· 0 citations
The Lightweight Action Inversion and Riemannian rectification network (LightAIR) is proposed, which introduces textual semantic priors as anchors via a lightweight action inversion operator to extract pure action features, thereby overcoming visual-inherent coupling.
Yulun Zhang, Zixu Li, Zhiwei Chen et al.· 0 citations
A cross-domain egocentric video question answering benchmark designed to evaluate whether multimodal large language models can generalize beyond common daily-life scenarios, and two official Codabench tracks.
Yu-Qian Fu, Tianwen Qian, Yanjun Li 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.