INDEX

Explanations

specific names and technical details in the text

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_fs12288-jb

Configuration

jbloom/GPT2-Small-Feature-Splitting-Experiment-Layer-8/blocks.8.hook_resid_pre_12288

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

Skylion007/openwebtext

Features

12,288

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

Fif

-0.81

 Piercing

-0.75

TY

-0.72

Squ

-0.70

Dou

-0.70

 cherry

-0.69

Spy

-0.68

 Jason

-0.68

Apr

-0.68

 bicy

-0.67

POSITIVE LOGITS

org

0.97

olution

0.90

iod

0.86

pool

0.84

ion

0.84

isons

0.83

ione

0.83

intern

0.82

wald

0.81

riot

0.81

Activations Density 0.360%

specific names and technical details in the text

No Comments

No Known Activations