Kognic launches Language Grounding for driving models
Kognic launched Language Grounding workflows for vision-language and vision-language-action driving models, adding structured natural-language reasoning data through write, edit and rank annotation modes.
Evidence notes
- The capability targets causal reasoning for autonomous-driving decisions.
- Workflows include original writing, editing model proposals and ranking outputs.
- Kognic also introduced a Chain of Causation method designed to limit hindsight bias.
Company context
Kognic provides annotation software and specialist data services for autonomous-driving and ADAS development. Its platform handles synchronized camera, LiDAR and radar data, supporting 3D labelling, sensor fusion, quality assurance and language-grounding workflows for perception and end-to-end models. Founded in Gothenburg in 2018, the company works with vehicle manufacturers, autonomy developers and major automotive suppliers.