Fengxiang He
PulseAugur coverage of Fengxiang He — every cluster mentioning Fengxiang He across labs, papers, and developer communities, ranked by signal.
-
LLM reasoning exhibits irrationality beyond value alignment, study finds
A new research paper from arXiv explores the concept of "rational value risk" in large language models, suggesting that even well-aligned models can exhibit irrationality during reasoning. This risk is quantified as a d…
-
PEAR protocol enhances LLM debate reliability with adaptive routing
Researchers have developed a new multi-agent debate protocol called PEAR (Permutation-Equivariant Adaptive Routing) to enhance the reliability of large language models. Unlike fixed topologies, PEAR dynamically reconfig…
-
DeXposure-FM: New Graph Foundation Model for DeFi Credit Exposure
Researchers have introduced DeXposure-FM, a novel time-series graph foundation model designed to quantify and forecast credit exposure within decentralized finance (DeFi) networks. This model aims to address the contagi…
-
Equivariant neural networks may limit expressive power, paper finds
A new research paper explores the trade-offs of using equivariant neural networks, which are designed to leverage data symmetries. The study, focusing on 2-layer ReLU networks, demonstrates that enforcing these symmetry…