Search papers, labs, and topics across Lattice.
University of California
7
0
12
Training LLMs without ground-truth solutions can yield significant performance improvements, as shown by RiVER's success in enhancing both score-based and exact-solution benchmarks.
Future-oriented reasoning in tabular question answering is now feasible with the introduction of a dataset and framework that outperforms existing LLM capabilities.
SimSD achieves a remarkable 7.46x increase in decoding throughput for diffusion language models without sacrificing generation quality.
Finally, voice agents can become interactive humanoid avatars: EchoAvatar generates coherent full-body motion from streaming audio, even switching between speech and music, all while being controllable by an LLM.
VLMs can be taught to self-correct hallucinations at the token level, leading to substantial gains in reasoning accuracy across diverse benchmarks.
SlideAgent not only updates presentation slides with natural language instructions but also preserves their original design, setting a new standard for automated slide management.
Today's best AI agents still fail more than half the time on real-world tasks combining vision, search, and coding, revealing critical gaps in reasoning and tool use.
