INDEX

Explanations

phrases related to causes and results

oai_token-act-pair · gpt-3.5-turbo

phrases indicating causality and consequences related to various events or actions

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

PlotsExplanationShow Test FieldDefault Test Text

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

è£

-0.82

issance

-0.76

NES

-0.74

thank

-0.73

Commission

-0.72

ulous

-0.72

adel

-0.71

Laughs

-0.69

istry

-0.69

icol

-0.68

POSITIVE LOGITS

 untreated

1.17

 uncontrolled

1.10

 ingest

1.08

 unchecked

1.01

 exposure

1.00

 improperly

0.98

 mishand

0.98

 negligence

0.97

 misuse

0.97

 improper

0.95

Activations Density 0.409%

phrases related to causes and results

phrases indicating causality and consequences related to various events or actions

No Comments

No Known Activations