INDEX

Explanations

text that has been edited or condensed

oai_token-act-pair · gpt-3.5-turbo

references to editing or altered content

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

falls

-0.79

fw

-0.76

aches

-0.74

fall

-0.74

ptoms

-0.74

gaard

-0.72

hood

-0.72

 Arcade

-0.72

phal

-0.69

fell

-0.69

POSITIVE LOGITS

 summ

0.84

 excerpts

0.82

 annex

0.78

 transcript

0.76

edited

0.72

 edited

0.68

icum

0.66

iola

0.66

 narration

0.65

ison

0.64

Activations Density 0.024%

text that has been edited or condensed

references to editing or altered content

No Comments

No Known Activations