INDEX

Explanations

references to news publication "Telegraph."

oai_token-act-pair · gpt-3.5-turbo

occurrences of the word "Telegraph."

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

regn

-0.82

perm

-0.78

etime

-0.74

vae

-0.72

ĸļ

-0.69

outine

-0.66

 tuna

-0.66

vasive

-0.66

venge

-0.65

razil

-0.64

POSITIVE LOGITS

 Telegraph

0.97

orial

0.86

legraph

0.84

ieth

0.83

shire

0.82

ulla

0.80

sburg

0.80

biz

0.79

 Newsp

0.78

 Editorial

0.76

Activations Density 0.007%

references to news publication "Telegraph."

occurrences of the word "Telegraph."

No Comments

No Known Activations