← 🏆 Competitions
Apply Now ↗
Closing Soon 🏆 Competitions
IJCAI 2026 CAR-bench — LLM Agent Reliability Challenge
CAR-bench evaluates LLM agents as automotive in-car voice assistants, testing multi-turn task completion, hallucination resistance, and disambiguation of ambiguous requests. Leaderboard submissions open through June 30, 2026.
Why it matters
Benchmark competition with direct TMLR publication pathway — one of the few competitions where participation itself produces a citable research output.
Application Deadline
June 30, 2026
20 days remaining
CAR-bench is an academic benchmark competition focused on LLM agent reliability in a constrained, safety-relevant domain: automotive in-car voice assistants.
What gets evaluated:
- Multi-turn task completion with tool and policy chaining
- Hallucination resistance when capabilities are missing
- Disambiguation of ambiguous user requests
- Interaction with an LLM-simulated user across a mutable environment
Two tracks:
- Open Track — any model or approach
- Cerebras Fast-Reasoning Track — optimised for low-latency inference
Timeline:
- Development phase + leaderboard: May–June 30, 2026
- Final submission deadline: July 15, 2026 (23:59 AoE)
Useful for teams working on agentic systems, tool use, or production LLM reliability.