INDEX

Explanations

references to the concept of removal or eliminating something

oai_token-act-pair · gpt-4o-mini Triggered by @bot

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 8-res_post_32k-oai

Configuration

jbloom/GPT2-Small-OAI-v5-32k-resid-post-SAEs/v5_32k_layer_8.pt

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

Skylion007/openwebtext

Features

32,768

Data Type

torch.float32

Hook Name

blocks.8.hook_resid_post

Hook Layer

Architecture

standard

Context Size

Dataset

Skylion007/openwebtext

Activation Function

topk

Embeds

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

£ı

-0.70

gerald

-0.70

 msec

-0.67

akening

-0.65

gregation

-0.63

enegger

-0.63

zag

-0.62

eers

-0.61

ingham

-0.61

ERAL

-0.60

POSITIVE LOGITS

uder

0.80

 foreskin

0.73

leased

0.71

 rubble

0.69

ased

0.67

ained

0.67

pert

0.66

uters

0.66

leted

0.66

 from

0.66

Activations Density 0.068%

references to the concept of removal or eliminating something

No Comments

No Known Activations