PPL
Ming Liu
Papers (7)
LiveBrowseComp: Are Search Agents Searching, or Just Verifying What They Already Know?Bandwidth-Efficient and Privacy-Preserving Edge-Cloud Many-to-Many Speech TranslationSelf-Training Doesn't Flatten Language -- It Restructures It: Surface Markers Amplify While Deep Syntax DiesThe Readout Shortcut: Positional Number Copying Dominates Arithmetic CoT Readout in Small Language ModelsIn-Context Fixation: When Demonstrated Labels Override Semantics in Few-Shot ClassificationMore Is Not Always Better: Cross-Component Interference in LLM Agent ScaffoldingDecodable but Not Corrected by Fixed Residual-Stream Linear Steering: Evidence from Medical LLM Failure Regimes