Search results

Enhance: Support training on different of activations between base and reward models.

Status: Open.

#15 In apartresearch/Interpreting-Learned-Feedback-Patterns;

· amirabdullah19852020 opened on Apr 7, 2024

More data: Create two datasets optimized for Vader, and upload to datasets

data

Status: Open.

#14 In apartresearch/Interpreting-Learned-Feedback-Patterns;

· amirabdullah19852020 opened on Apr 7, 2024

More reward models: Train a DPO model and autoencoders for unalignment dataset.

model

Status: Open.

#13 In apartresearch/Interpreting-Learned-Feedback-Patterns;

· amirabdullah19852020 opened on Apr 7, 2024

More reward models: Create contrastive pairs dataset for unalignment dataset.

data

Status: Open.

#12 In apartresearch/Interpreting-Learned-Feedback-Patterns;

· amirabdullah19852020 opened on Apr 7, 2024

More reward models: Create a contrastive pairs dataset for helpful / harmless

data

Status: Open.

#11 In apartresearch/Interpreting-Learned-Feedback-Patterns;

· amirabdullah19852020 opened on Apr 7, 2024

More reward models: Train a DPO model and autoencoders for hh-hrlf

model

Status: Open.

#10 In apartresearch/Interpreting-Learned-Feedback-Patterns;

· amirabdullah19852020 opened on Apr 7, 2024

Enhance: Integrate with SAELens for the sparse autoencoder training.

enhancementNew feature or requestNew feature or request

Status: Open.

#9 In apartresearch/Interpreting-Learned-Feedback-Patterns;

· amirabdullah19852020 opened on Apr 7, 2024