INDEX

Explanations

phrases or words that involve emphasis or importance

oai_token-act-pair · gpt-3.5-turbo

the contraction "t" as in "can't" or "won't."

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 0-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.0.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.0.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

 mascul

-0.62

 domestically

-0.57

ADRA

-0.55

 heels

-0.52

 validated

-0.52

 federation

-0.52

 Reaper

-0.52

 desk

-0.51

 aside

-0.51

 sparing

-0.51

POSITIVE LOGITS

3.69

ti

1.99

tan

1.95

tin

1.94

tif

1.88

tz

1.85

ta

1.82

ts

1.79

tal

1.79

tor

1.77

Activations Density 0.161%

phrases or words that involve emphasis or importance

the contraction "t" as in "can't" or "won't."

No Comments

No Known Activations