INDEX

Explanations

action words referring to observing or experiencing entertainment content

oai_token-act-pair · gpt-3.5-turbo

references to watching and listening activities

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 5-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.5.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.5.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

 Berm

-0.78

 ransom

-0.72

 deployed

-0.66

 dispatch

-0.65

 exoner

-0.65

 operational

-0.65

staking

-0.63

nel

-0.63

 sworn

-0.62

 regener

-0.62

POSITIVE LOGITS

ables

0.79

 enjoyment

0.79

artifacts

0.76

 Trails

0.74

 Watching

0.72

ences

0.70

é¾įå

0.69

 habits

0.69

ERS

0.69

estyles

0.69

Activations Density 0.177%

action words referring to observing or experiencing entertainment content

references to watching and listening activities

No Comments

No Known Activations