INDEX

Explanations

dates and locations in news articles

oai_token-act-pair · gpt-3.5-turbo

punctuation marks, particularly periods and commas

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 8-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.8.hook_resid_pre

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

oris

-0.68

NCT

-0.66

Output

-0.66

ogen

-0.63

verbs

-0.63

IME

-0.61

olate

-0.60

xia

-0.60

ãĥ¼ãĥĨãĤ£

-0.59

imus

-0.59

POSITIVE LOGITS

uits

0.84

imentary

0.72

taboola

0.65

 Boone

0.63

DAQ

0.63

nesday

0.62

ahoo

0.62

atri

0.61

lication

0.61

 Bethlehem

0.61

Activations Density 0.042%

dates and locations in news articles

punctuation marks, particularly periods and commas

No Comments

No Known Activations