INDEX

Explanations

words related to deeply rooted ideas or positions

oai_token-act-pair · gpt-3.5-turbo

terms related to established or deep-rooted interests and their influence

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 1-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.1.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.1.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

hner

-0.86

oping

-0.84

ynthesis

-0.82

othy

-0.76

kers

-0.75

ioxide

-0.74

asers

-0.74

paces

-0.73

umbn

-0.72

ionic

-0.72

POSITIVE LOGITS

 entrenched

1.31

 ingrained

1.08

 vested

0.83

 behavi

0.80

 embroiled

0.78

 incumbent

0.77

 incumb

0.76

 proble

0.75

 entangled

0.75

lishes

0.74

Activations Density 0.015%

words related to deeply rooted ideas or positions

terms related to established or deep-rooted interests and their influence

No Comments

No Known Activations