INDEX

Explanations

words related to news articles, often with a sense of continuation or following on from a previous topic

oai_token-act-pair · gpt-3.5-turbo

instances of the word "Next."

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 10-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.10.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.10.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

ans

-0.68

hips

-0.65

ocker

-0.64

lees

-0.63

kay

-0.63

tics

-0.61

 Feldman

-0.60

hess

-0.59

hed

-0.58

ux

-0.58

POSITIVE LOGITS

door

1.06

 Steps

1.02

STEP

0.93

 steps

0.92

 generation

0.85

Gen

0.84

 door

0.83

 installment

0.82

 Generation

0.81

 step

0.81

Activations Density 0.032%

words related to news articles, often with a sense of continuation or following on from a previous topic

instances of the word "Next."

No Comments

No Known Activations