Emotion Probes Crash From 82% to 5% Without Keywords

Also Worth Noting

• The Collapse of Heterogeneity in Silicon Philosophers score 4
  机构: Stanford; 关键词(3): fine-tuning, DPO, open-source
• RouteNLP: Closed-Loop LLM Routing with Conformal Cascading and Distillation Co-Optimization score 4
  关键词(4): distillation, deployment, serving, latency; 顶会接收: ACL
• AgentEval: DAG-Structured Step-Level Evaluation for Agentic Workflows with Error Propagation Tracking score 4
  关键词(4): production, agentic, tool use, reasoning; 顶会接收: ACL
• ComplianceNLP: Knowledge-Graph-Augmented RAG for Multi-Framework Regulatory Gap Detection score 4
  关键词(3): distillation, deployment, RAG; 顶会接收: ACL
• FinGround: Detecting and Grounding Financial Hallucinations via Atomic Claim Verification score 4
  关键词(3): deployment, latency, RAG; 顶会接收: ACL
• Quasi-Equivariant Metanetworks score 4
  关键词(1): reasoning; 顶会接收: ICLR
• S2G-RAG: Structured Sufficiency and Gap Judging for Iterative Retrieval-Augmented QA score 4
  关键词(4): lightweight, retrieval-augmented, RAG, reasoning; 顶会接收: ACL
• Knowledge Vector of Logical Reasoning in Large Language Models score 4
  关键词(1): reasoning; 顶会接收: ACL
• MTRouter: Cost-Aware Multi-Turn LLM Routing with History-Model Joint Embeddings score 3
  顶会接收: ACL
• AIPsy-Affect: A Keyword-Free Clinical Stimulus Battery for Mechanistic Interpretability of Emotion in Language Models score 3
  机构: MIT