TanmaySah commited on
Commit
fe90e85
·
verified ·
1 Parent(s): 926dbf1

docs: add Model Card | GitHub navigation header

Browse files
Files changed (1) hide show
  1. README.md +4 -6
README.md CHANGED
@@ -12,14 +12,12 @@ short_description: Order-Equivariant Probabilistic Decision Model
12
 
13
  # InvariantOne — Permutation-Equivariant Decision Model Demo
14
 
15
- Interactive demonstration of **InvariantOne-v1**, an order-equivariant probabilistic decision model for runtime multiple-choice action selection.
16
 
17
- - **Hugging Face Model:** [TanmaySah/InvariantOne-v1](https://huggingface.co/TanmaySah/InvariantOne-v1)
18
- - **GitHub Repository:** [tradertanmay/InvariantOne](https://github.com/tradertanmay/InvariantOne)
19
- - **PyPI Package:** [`invariantone==1.0.2`](https://pypi.org/project/invariantone/)
20
 
21
  ### Key Characteristics
22
  - **Candidate-Independent Scoring:** Each candidate option is encoded independently with context, eliminating position and option-order biases.
23
  - **Analytic Distributional Equivariance:** Exact permutation equivariance holds by construction in real arithmetic.
24
- - **Calibrated Softmax Readout:** Outputs probabilities normalized by frozen calibration temperature $\tau^* = 1.0091$.
25
- - **Dynamic Options ($K \ge 2$):** Supports arbitrary numbers of runtime candidate actions (empirical benchmarks evaluated on $K=4$).
 
12
 
13
  # InvariantOne — Permutation-Equivariant Decision Model Demo
14
 
15
+ [Model Card](https://huggingface.co/TanmaySah/InvariantOne-v1) | [GitHub](https://github.com/tradertanmay/InvariantOne)
16
 
17
+ Interactive demonstration of **InvariantOne-v1**, an order-equivariant probabilistic decision model for runtime multiple-choice action selection.
 
 
18
 
19
  ### Key Characteristics
20
  - **Candidate-Independent Scoring:** Each candidate option is encoded independently with context, eliminating position and option-order biases.
21
  - **Analytic Distributional Equivariance:** Exact permutation equivariance holds by construction in real arithmetic.
22
+ - **Calibrated Softmax Readout:** Outputs probabilities normalized by frozen calibration temperature τ* = 1.0091.
23
+ - **Dynamic Options (K ≥ 2):** Supports arbitrary numbers of runtime candidate actions (empirical benchmarks evaluated on K = 4).