inclusionAI Releases LemonDecider 4B v3 Decision Model
Start here
inclusionAI released LemonDecider 4B v3, a decision model fine-tuned from Qwen3.5-4B that returns choice, yes/no or ordinal probabilities in a single joint forward pass rather than generating answer tokens. The company says it trained the model on about 1.20M decision/QA rows; it scored 61.67 on the public Decision Index 0.4 suite, with median request latency of 49.9 ms. The released weights are a uniform FP32 average of four checkpoints from the same supervised fine-tuning run, followed by scalar temperature calibration — not a combination of two independent training runs. This training run is text-only with the vision tower frozen, and the HTTP bridge is also text-only. The model is published under Apache-2.0, with its code under MIT.
Generated from reports · updated 13 h ago
Report timeline
Follow the reports to see the event from different sides.
- inclusionAI ModelsinclusionAI Releases LemonDecider 4B v3 Decision Model Based on Qwen3.5-4B
inclusionAI released LemonDecider 4B v3, a decision model fine-tuned from Qwen3.5-4B that returns choice, yes/no or ordinal probabilities via a single joint forward pass instead of generating answer tokens. It was trained on about 1.20M decision/QA rows and scored 61.67 on the public Decision Index 0.4 suite, with median request latency of 49.9 ms.
Heat for this event
Not enough continuous observations yet to draw a trend.