INDEX

Explanations

details related to news articles or reports

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_fs768-jb

Configuration

jbloom/GPT2-Small-Feature-Splitting-Experiment-Layer-8/blocks.8.hook_resid_pre_768

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

Skylion007/openwebtext

Features

768

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

worldly

-0.63

oir

-0.62

ãĥ¥

-0.61

(>

-0.58

ARCH

-0.57

OND

-0.56

ente

-0.56

irection

-0.55

LF

-0.55

Actor

-0.54

POSITIVE LOGITS

 respectively

0.83

udeb

0.68

etc

0.65

0.64

 namely

0.63

 albeit

0.62

luaj

0.61

 disg

0.61

 disclaim

0.61

 which

0.60

Activations Density 3.041%

details related to news articles or reports

No Comments

No Known Activations