README update, markdown fix in Incentive.md

README.md

@@ -83,12 +83,7 @@ To address this growing challenge, SN34 aims to create the most accurate fully-g
 
 **Challenge Generation and Scoring (Blue Arrows)**
 
-For each challenge, the validator performs the following steps:
-1. Randomly samples a real or synthetic image/video from the cache
-2. Applies random augmentations to the sampled media
-3. Distributes the augmented data to 50 randomly selected miners for classification
-4. Updates its score vector based on each miner's historical performance and computed rewards for the current challenge
-5. Logs comprehensive challenge results to [Weights and Biases](https://wandb.ai/bitmindai/bitmind-subnet), including the generated media, original prompt, miner responses and rewards, and other challenge metadata
+For each challenge, the validator randomly samples a real or synthetic image/video from the cache, applies random augmentations to the sampled media, and distributes the augmented data to 50 randomly selected miners for classification. It then scores the miners responses, and logs comprehensive challenge results to [Weights and Biases](https://wandb.ai/bitmindai/bitmind-subnet), including the generated media, original prompt, miner responses and rewards, and other challenge metadata.
 
 **Synthetic Data Generation (Pink Arrows)**:
 

docs/Incentive.md

@@ -23,23 +23,24 @@ Validators keep track of miner performance using a score vector, which is update
 </em></p>
 
 
-## Longer Version
 
-### Rewards
+## Rewards
 > Total rewards are a weighted combination of video and image rewards. Rewards for both image and video challenges are the [Matthews Correlation Coefficient (MCC)](https://en.wikipedia.org/wiki/Phi_coefficient) of (up to) a miner's last 100 predictions, combined with the accuracy of their last 10. 
 
 Total rewards
+
 $$ 
 C_{total} = 0.6 \cdot C_{image} + 0.4 \cdot C_{video} 
 $$
 
 Rewards for modality *m*
+
 $$ 
-C_{m} = 0.5 \cdot MCC_{m} + 0.5 \cdot Accuracy_{m}
+C_m = 0.5 \cdot MCC_m + 0.5 \cdot Accuracy_m
 $$
 
 
-### Scores
+## Scores
 
 >Validators set weights based on historical miner performances, tracked by their score vector. 
 
@@ -52,7 +53,7 @@ $$
 A low *&alpha;* value places emphasis on a miner's historical performance, adding additional smoothing to avoid having a single prediction cause significant score fluctuations.
 
 
-### Weights
+## Weights
 
 > Validators set weights around once per tempo (360 blocks) by sending a normalized score vector to the Bittensor blockchain (in `UINT16` representation).
 
@@ -61,7 +62,7 @@ Weight normalization by L1 norm:
 $$w = \frac{\text{V}}{\lVert\text{V}\rVert_1}$$
 
 
-### Incentives
+## Incentives
 > The [Yuma Consensus algorithm](https://docs.bittensor.com/yuma-consensus) translates the weight matrix *W* into incentives for the subnet miners and dividends for the subnet validators
 
 Specifically, for each miner *j*, incentive is a function of rank *R*: