INDEX

Explanations

phrases indicating a specific course or trend

oai_token-act-pair · gpt-3.5-turbo

references to "direction" in various contexts

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

tein

-0.81

zb

-0.72

aughs

-0.65

bians

-0.62

Prof

-0.61

oak

-0.61

tk

-0.61

bley

-0.61

bery

-0.60

LU

-0.60

POSITIVE LOGITS

 direction

1.25

 directions

1.12

naire

0.89

arity

0.88

eering

0.85

direction

0.81

ality

0.81

finding

0.79

 Direction

0.79

forward

0.74

Activations Density 0.011%

phrases indicating a specific course or trend

references to "direction" in various contexts

No Comments

No Known Activations