INDEX

Explanations

instances where actions that were explicitly mentioned were done

oai_token-act-pair · gpt-3.5-turbo Triggered by @bot

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 8-res_fs24576-jb

Configuration

jbloom/GPT2-Small-Feature-Splitting-Experiment-Layer-8/blocks.8.hook_resid_pre_24576

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

Skylion007/openwebtext

Features

24,576

Data Type

torch.float32

Hook Point

blocks.8.hook_resid_pre

Architecture

standard

Context Size

128

Dataset

Skylion007/openwebtext

Hook Point Layer

Activation Function

relu

Embeds

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

heter

-0.74

isu

-0.71

cedented

-0.69

rooms

-0.69

houses

-0.67

asant

-0.65

stood

-0.65

objects

-0.64

arters

-0.62

mare

-0.61

POSITIVE LOGITS

actic

1.03

riks

0.71

iotic

0.71

netflix

0.69

nt

0.67

act

0.66

 happen

0.66

 offend

0.65

Finish

0.64

Activations Density 0.035%

instances where actions that were explicitly mentioned were done

No Comments

No Known Activations