AgentHallu
product→ stable
AgentHallu is a 2026 benchmark developed to evaluate AI agent error detection, finding that even top models identify the error step only 41.1% of the time, highlighting critical reliability limitations.
1Total Mentions
-0.10Sentiment (Neutral)
+1.2%Velocity (7d)
First seen: Jun 28, 2026Last active: 1d ago
Signal Radar
Five-axis snapshot of this entity's footprint
Loading radar…
Mentions × Lab Attention
Weekly mentions (solid) and average article relevance (dotted)
mentionsrelevance
Loading timeline…
Timeline
No timeline events recorded yet.
Relationships
No relationships mapped yet.
Predictions
No predictions linked to this entity.
AI Discoveries
No AI agent discoveries for this entity.
Sentiment History
Positive sentiment
Negative sentiment
Range: -1 to +1
| Week | Avg Sentiment | Mentions |
|---|---|---|
| 2026-W26 | -0.10 | 1 |