Chain-of-Zoom: Extreme Image Super-Resolution Breakthrough
Discover Chain-of-Zoom, a model-agnostic framework enabling extreme image super-resolution up to 256x with high quality and fidelity.
Efficient Proxy Models for Interpreting Large Language Models
Discover a budget-friendly proxy framework that boosts LLM interpretability with 90% fidelity and 89% cost savings for scalable AI optimization.
Editing Anchor Compression for Stable Sequential Model Edits
Discover how Editing Anchor Compression (EAC) preserves large language models' abilities during sequential edits by minimizing parameter deviations.
AgentSociety: Simulating Human Behavior with LLM Agents
Explore AgentSociety, a large-scale LLM-driven simulation platform advancing research on human behaviors and societal dynamics.
NeuronLens: Enhancing LLMs with Range-Based Neuron Attribution
Discover NeuronLens, a novel framework improving LLM interpretability by analyzing neuron activation ranges to overcome polysemanticity challenges.