task difficulty
An assessment of how complex or challenging a given problem is for an AI system, often influenced by factors such as dataset variability or specific goals.
- ARM: Adaptive Reasoning Model
- BountyBench: Dollar Impact of AI Agent Attackers and Defenders on Real-World Cybersecurity Systems
- The Overthinker's DIET: Cutting Token Calories with DIfficulty-AwarE Training
- Topology of Reasoning: Understanding Large Reasoning Models through Reasoning Graph Properties
- What Expressivity Theory Misses: Message Passing Complexity for GNNs
- When No Paths Lead to Rome: Benchmarking Systematic Neural Relational Reasoning
- Which Data Attributes Stimulate Math and Code Reasoning? An Investigation via Influence Functions