Robots Atlas>ROBOTS ATLAS
SII-Decision-27B

SII-Decision-27B

27B · Family: SII-Decision
Open-weight 27B decision model (SII-Decision family) producing structured JEVal decisions in a single forward pass via First-Token Decision Readout.
🔬 Research🔬 Research only⚖ Open sourceLLMSpecialized AI📁 SII-Decision
Parameters
27B
parameters
Access:DownloadDeployment:💻 Local

Overview

SII-Decision-27B is an open-weight 27-billion-parameter decision model from the SII-Decision family. It is a fine-tune of Qwen3.8-27B, built for fast, structured decisions in the JEVal format: single-option choice, binary judgment and score assignment.

The model uses the First-Token Decision Readout method, which constrains the next-token probability distribution of an open-weight LLM to the allowed option labels only and returns a decision in a single forward pass, without unrolling full reasoning.

Training relies on Reasoning-to-Readout Self-Distillation: the readout learns to match the distribution the model arrives at after reasoning, without gold labels and without a stronger teacher model. Full-weight and LoRA variants are available.

On the JEVal benchmark, the 27B variant reaches 79.24% accuracy, 0.0149 ECE (the lowest calibration error among the 25 tested configurations) and a 0.279 raw Brier score. In the SocioBench social simulation it attains 39.87% accuracy at 8.36 s per item, and in ElectionSim it correctly calls 47 of 51 states and 12 of 15 battleground states.

The model was developed by the Shanghai Innovation Institute (SII) together with Fudan University, Tongji University, King's College London and the Chinese University of Hong Kong. Weights and code are released under the Apache-2.0 license (inherited from the base Qwen weights).

Classification
LLMSpecialized AI
Family: SII-Decision
Access & deployment
Download
Local
Weights: Open source
Key parameters
🧩 Parameters: 27B
📥 Input: text

Technical specification

Parameters
27B
parameters
License
Apache-2.0
Modalities
⬇ Input
text
⬆ Output
structured_datatext

Capabilities and applications

Native model capabilities
Decision making
The ability to choose the best option from a defined set of possibilities based on input state, usually with a probability estimate for each option.
Category: reasoning
Classification
Assigning an observation to one of predefined classes (binary or multi-class). Output: class label and optionally probabilities.
Category: other
Structured output
Producing data in structured formats such as JSON.
Category: structured_generation
Instruction following
Precisely following instructions contained in the prompt: response format, length, style, constraints (e.g. 'reply in six words'). GPT-5.1 significantly improved this capability compared to GPT-5.
Category: language
Reasoning
The model's ability to reason logically and solve complex problems.
Category: reasoning

Benchmark results

5 benchmarks
JEVal
accuracy · First-Token Decision Readout; SII-Decision-27B variant
79.24%
📄 paper
JEVal
ECE · Lowest Expected Calibration Error among 25 configurations; 27B variant
0.0149
📄 paper
JEVal
Brier score (raw) · Raw Brier score; 27B variant
0.279
📄 paper
SocioBench
accuracy · Social simulation; 8.36 s per item; 27B variant
39.87%
📄 paper
ElectionSim
poprawne stany · Correctly called states (12/15 battleground); 27B variant
47/51
📄 paper

Technical architecture

Core Architecture
Model Form