Large language models (LLMs) often solve a task when all instructions are given in a single prompt, but fail when the same information is revealed gradually across turns. When a clean FULL prompt and a RAW-SHARDED conversation contain the same complete user evidence, the model should still arrive at the same answer. We...
Zizhuo Lin, Quanling Liu, Jinsheng Quan et al.· 0 citations
Large language models are increasingly asked to stand in for particular people, and interviewing the person is the natural way to tell a model who they are. We test what a one-hour LLM-led adaptive interview tells a model about an individual's value-laden decisions. 46 adults completed such an interview and then answer...
As distributed model training scales to span hundreds of thousands of GPUs, scale-out networks face unprecedented performance and efficiency demands. NVIDIA Spectrum-X Ethernet has been designed from the ground up to achieve predictable and stable network performance with high utilization and low latency. This paper pr...
Sajy Khashab, Albert Gran Alcoz, Alon Gal et al.· 0 citations
This work proposes DEFLECT, an offline post-training framework that turns this temporal contrast into preference supervision for a policy acting from stale inputs and improves delay robustness across the evaluated VLA policies and settings.
Yi-Xiang Zhu, Yong-Hao Chen, Jia-Xiang Zou et al.· 3 citations
Reach audiences
Advertise in front of researchers, engineers, and readers.
DiDi-Merging is proposed, which jointly allocates low-rank capacity to these components through differentiable rank selection, followed by data-free parameter-space refinement, followed by data-free parameter-space refinement.
CounterCount goes beyond showing that VLMs fall back on knowledge priors, revealing that once a familiar context is recognized, models favor prior knowledge over available visual evidence rather than failing to encode that evidence.
Reem Alzahrani, Bushra Bin Hemid, Hassan Alshanqiti et al.· 0 citations
DRAPE (Dynamic Cross-Modal Prompt Generation), a prompt-learning framework that synthesizes continuous instance-specific soft prompts for MCIT, and achieves state-of-the-art performance among representative prompt-based and LoRA-based continual-learning baselines.
A bottom-up cascaded knowledge distillation (CKD) framework is proposed, inspired by human formal education systems, and introduces one (potentially, more) additional Teacher(s) of intermediate capacity that gradually bring the Student network to the next level, where the next (higher-capacity) Teacher can take over.
Standard LLM-based speech recognition systems process utterances in isolation, limiting their ability to leverage conversational context. Here we study whether multimodal context from prior turns improves LLM-based ASR and how to represent that context efficiently. We find that, after supervised multi-turn training, co...
Shashi Kumar, Esa\'u Villatoro-Tello, Sergio Burdisso et al.· 0 citations
These findings show that lightweight additive forecasting strategies can remain highly competitive for urban air-quality prediction, offering a practical balance between accuracy, interpretability, and computational efficiency.
Moazzam Umer Gondal, Hamad Ul Qudous, Asma Ahmad Farhan et al.· 0 citations
Current alignment evaluation mostly measures whether models encode dangerous concepts and whether they refuse harmful requests. Both miss the layer where alignment often operates: routing from concept detection to behavioral policy. We study political censorship in Chinese-origin language models as a natural experiment...
This work gives semantic and temporal guidance from the keyframes and text onto each intermediate frame through Keyframe-anchored Attention Bias and better enforce frame consistency with Rescaled Temporal RoPE, which allows self-attention to attend to keyframes more faithfully.
Tae Eun Choi, Sumin Shim, Junhyeok Kim et al.· arXiv.org· 3 citations
