The Variational JEPA (Var-JEPA), which makes the latent generative structure explicit by optimizing a single Evidence Lower Bound (ELBO) and yields meaningful representations without ad-hoc anti-collapse regularizers and allows principled uncertainty quantification in the latent space.
Moritz Gögl, Christopher Yau· arXiv.org· 3 citations· ⚡1
These findings demonstrate that current defense paradigms optimize for single-turn refusal benchmarks while rendering multi-step agents fundamentally unreliable, necessitating new approaches that preserve tool execution competence under adversarial conditions.
A consistency boundary analysis is presented that characterizes when diagonal short-memory SSMs can approximate causal attention and identifies structural gaps that remain and proposes InfoMamba, an attention-free hybrid architecture that consistently outperforms strong Transformer and SSM baselines.
Youjin Wang, Jiaqi Zhao, Rong Fu et al.· arXiv.org· 0 citations
There is potential to improve cross-lingual parametric knowledge transfer during post-training by providing the LLMs with the key entities of the questions in their source language and finding that this disproportionately improves cross-script questions.
Lucas Bandarkar, Alan Ansell, Trevor Cohn· arXiv.org· 3 citations
Reach audiences
Advertise in front of researchers, engineers, and readers.
The results show that improving IF in LRMs can significantly enhance privacy, suggesting a promising direction for future privacy-aware LRMs, and introduces an SFT dataset that teaches models to follow general instructions throughout their reasoning process.
Haritz Puerto, Haonan Li, Xudong Han et al.· 0 citations
FENCE, a bilingual (Korean-English) multimodal dataset for training and evaluating jailbreak detectors in financial applications, provides a focused resource for advancing multimodal jailbreak detection in finance and for supporting safer, more reliable AI systems in sensitive domains.
Mirae Kim, Seonghun Jeong, Youngjun Kwak· arXiv.org· 0 citations
Activation Steering Adapter (ASA), a training-free, inference-time controller that performs a single-shot mid-layer intervention and targets tool domains via a router-conditioned mixture of steering vectors with a probe-guided signed gate to amplify true intent while suppressing spurious triggers is proposed.
Youjin Wang, Run Zhou, Rong Fu et al.· 4 citations· ⚡2
Experiments on simulated and real-world benchmarks demonstrate that SCALE improves state-of-the-art VLAs and outperforms existing TTS methods while maintaining single-pass efficiency.
Hyeonbeom Choi, Daechul Ahn, Youhan Lee et al.· arXiv.org· 3 citations
This work proposes a cognitive-inspired, multi-agent framework that operationalizes Conceptual Blending Theory (CBT) through a novel Schema Grammar ("G"), providing a rigorous foundation for cross-domain logic re-instantiation.
Yu Xu, Yuxin Zhang, Juan Cao et al.· arXiv.org· 4 citations
The architecture family implementing this function class is named CoFrGeNets - Continued Fraction Generative Networks, and novel architectural components based on this function class that can replace Multi-head Attention and Feed-Forward Networks in Transformer blocks while requiring much fewer parameters are designed.
Amit Dhurandhar, Vijil Chenthamarakshan, Dennis Wei et al.· arXiv.org· 0 citations
WorldMind is introduced, a framework that autonomously constructs a symbolic World Knowledge Repository by synthesizing environmental feedback that unifies Process Experience to enforce physical feasibility via prediction errors and Goal Experience to guide task optimality through successful trajectories.
Baochang Ren, Yunzhi Yao, Rui Sun et al.· arXiv.org· 3 citations· ⚡1
Experimental results demonstrate that FastSLM achieves competitive performance across diverse speech-language tasks while requiring substantially fewer speech tokens and FLOPs than existing speech-language models.
