VISTA: A Visual Harness for Reasoning in an Interactive World
arXiv cs.AI · 2 Oct 2026, 03:59 AEST
We show that multimodal models possess strong reasoning abilities and that an appropriate harness can unlock their potential to solve tasks across diverse interactive environments. We introduce VISTA, a visual harness that gives a general-purpose multimodal model long-horizon vision.
Key facts
- We show that multimodal models possess strong reasoning abilities and that an appropriate harness can unlock their potential to solve tasks across diverse interactive environments.
- We introduce VISTA, a visual harness that gives a general-purpose multimodal model long-horizon vision.
- VISTA allows the model to directly perceive the environment through visual observations and maintains a lossless visual memory that preserves past observations in their original form.
- The model can actively retrieve these observations and reorganize its visual input as it reasons.
- On ARC-AGI-3, VISTA improves Claude Opus 5.0's Relative Human Action Efficiency score from 40.68 to a perfect 100.00, with the model completing all 25 public games using 57.4% fewer actions than first-time human.
Read aloud
We show that multimodal models possess strong reasoning abilities and that an appropriate harness can unlock their potential to solve tasks across diverse interactive environments. We introduce VISTA, a visual harness that gives a general-purpose multimodal model long-horizon vision. We show that multimodal models possess strong reasoning abilities and that an appropriate harness can unlock their potential to solve tasks across diverse interactive environments. We introduce VISTA, a visual harness that gives a general-purpose multimodal model long-horizon vision. VISTA allows the model to directly perceive the environment through visual observations and maintains a lossless visual memory that preserves past observations in their original form. The model can actively retrieve these observations and reorganize its visual input as it reasons. On ARC-AGI-three, VISTA improves Claude Opus five point zero's Relative Human Action Efficiency score from forty point six eight to a perfect one hundred point zero zero, with the model completing all twenty five public games using fifty seven point four percent fewer actions than first-time human.
Sources in this cluster
- VISTA: A Visual Harness for Reasoning in an Interactive World
arXiv cs.AI · 2 Oct 2026, 03:59 AEST