JudgeMyAISEMA-Auto: Bridging the Gap in Multi-Turn Jailbreak Attackers
SEMA-Auto advances multi-turn jailbreak attacks with autonomy and efficiency, outperforming static and interactive baselines.
GraphDancer: How Curriculum Reinforcement Learning Unlocks Cross-Domain Graph Reasoning in LLMs
GraphDancer leverages curriculum reinforcement learning to train LLMs for cross-domain graph reasoning.
Compaction Strategies in Long Horizon Agentic Tasks: A Comprehensive Analysis
Explore the impact of compaction strategies on long horizon agentic tasks with frontier models.
A-MapReduce: How Agentic MapReduce Achieves State-of-the-Art Wide Search Performance
A-MapReduce revolutionizes wide search with agentic MapReduce, delivering 5.11% to 17.50% F1 improvements and 45.8% faster execution.
SOPO: Socratic Guided Policy Optimization for Span-Level Hallucination Detection
SOPO: Socratic Guided Policy Optimization for Span-Level Hallucination Detection
Ground Truth Sensitivity: How Annotation Coverage Shapes Retrieval-Augmented Generation (RAG) Evaluation
Exploring how ground truth annotation coverage impacts retrieval scores in RAG systems for breast lesion management.
UniGuard Improves Cross-Interface Guardrail Generalization by 45.12 F1 Points
UniGuard improves cross-interface guardrail generalization by 45.12 F1 points. Learn how multi-view training enhances safety.
ATTUNE: Agentic Throughput Tuning for LLM Training
ATTUNE agentic throughput tuning for LLM training raises throughput by 1.7x on average.
Quant-Arena: Benchmarking LLM Agents for Quantitative Trading Research
Quant-Arena benchmarks LLM agents for quantitative trading research, evaluating end-to-end strategy development and execution.
Quant-Arena Benchmark Reveals LLMs Can’t Beat Equal-Weight Market in 2024 Quantitative Research
Quant-Arena benchmark reveals LLMs struggle to beat equal-weight market in 2024 quantitative research.