Darwin-27B-ZTC-v2
An open-source decision model with 27 billion parameters released by VIDRAFT. It employs the Zero-Token Calibration (ZTC) method and is designed specifically for decision nodes in agent graphs. The model supports structured probabilistic output and serves as an open-source alternative to proprietary typed decision models such as TypeSafe AI Jev.
Timeline (last 90 days)
VIDRAFT released an open-weight model Darwin-27B-ZTC-v2 under the Apache-2.0 license.
Darwin-27B-ZTC-v2 ranked 1st out of 102 models on the System One Mosaic Benchmark decision engine leaderboard.
Darwin-27B-ZTC-v2 achieved a Borda score of 89.58 and a task average score of 66.46.
Darwin-27B-ZTC-v2 uses a method called Zero-Token Calibration (ZTC).
The original article does not disclose the Hugging Face repository address, GitHub link, or API endpoint for Darwin-27B-ZTC-v2.
VIDRAFT Darwin-27B-ZTC-v2 is an open-weight model that uses ZTC (Zero-Token Confidence) to score decisions, generating no tokens during the decision process.
All Facts (6)
Darwin-27B-ZTC-v2 ranked 1st out of 102 models on the System One Mosaic Benchmark decision engine leaderboard.
50%UnverifiedDarwin-27B-ZTC-v2 achieved a Borda score of 89.58 and a task average score of 66.46.
50%UnverifiedDarwin-27B-ZTC-v2 uses a method called Zero-Token Calibration (ZTC).
50%UnverifiedThe original article does not disclose the Hugging Face repository address, GitHub link, or API endpoint for Darwin-27B-ZTC-v2.
50%UnverifiedVIDRAFT Darwin-27B-ZTC-v2 is an open-weight model that uses ZTC (Zero-Token Confidence) to score decisions, generating no tokens during the decision process.
50%UnverifiedVIDRAFT released an open-weight model Darwin-27B-ZTC-v2 under the Apache-2.0 license.
50%