Search papers, labs, and topics across Lattice.
11
0
14
0
This work proposes DualViewEval, an agent benchmark compression method that jointly exploits outcome and process relations to learn an exact-size miniset and predict the full-benchmark scores, and achieves the best results in all datasets.
This work presents a tail-risk-aware turn release scheduling method that jointly decides which ready turn to release next and how much released but unfinished work to maintain, and reduces the P95 of workflow flow time under contention, achieving up to a \(3.50\times\) speedup.
Models that appear equally accurate can possess vastly different agentic reasoning capabilities, revealing the hidden complexities of LLM performance.
Latent-OPD reveals that distilling latent representations at trajectory endpoints can dramatically enhance video reasoning efficiency in LMMs.
Cross-path reasoning reveals untapped research ideas, outperforming traditional methods and rivaling human insights in scholarly evolution.
Video generators may convincingly simulate real-world dynamics, but they often fail to understand the underlying causal relationships, revealing a critical gap in their reasoning capabilities.
TopoAgent's innovative use of dynamic graph evolution allows for noise-resistant scientific reasoning that outperforms traditional linear models.
LLMs are evolving from reactive chatbots to proactive digital colleagues, fundamentally changing how AI can assist in complex tasks.
Forget bolting vision onto language models – truly powerful multimodal AI demands rethinking architectures from the ground up.
Humanoids can now play ping-pong using *only* onboard cameras, pulling off whole-body smashes and crouch shots with impressive agility.
