INDEX

Explanations

words related to rankings, contests, and notable achievements

oai_token-act-pair · gpt-3.5-turbo

suffixes or fragments commonly found in longer words or terms

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 3-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.3.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.3.hook_resid_pre

Architecture

standard

Context Size

128

Dataset

Skylion007/openwebtext

Hook Point Layer

Activation Function

relu

Embeds

PlotsExplanationShow Test FieldDefault Test Text

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

etheless

-0.93

udeb

-0.80

 challeng

-0.79

 neut

-0.79

 seiz

-0.78

 carbohyd

-0.77

sembly

-0.76

 nursing

-0.72

perty

-0.72

 citiz

-0.70

POSITIVE LOGITS

Gate

0.84

Kings

0.82

 Kings

0.81

Knight

0.79

frog

0.76

croft

0.76

 Jamaica

0.76

 Logic

0.76

joy

0.75

 Jackets

0.75

Activations Density 0.183%

words related to rankings, contests, and notable achievements

suffixes or fragments commonly found in longer words or terms

No Comments

No Known Activations