INDEX

Explanations

mentions of specific names, likely related to a news story or event

oai_token-act-pair · gpt-3.5-turbo

proper nouns related to people, likely focusing on names and roles

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 3-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.3.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.3.hook_resid_pre

Architecture

standard

Context Size

128

Dataset

Skylion007/openwebtext

Hook Point Layer

Activation Function

relu

Embeds

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

ĳ

-0.65

orses

-0.63

ccording

-0.61

iggurat

-0.61

rises

-0.61

utes

-0.60

olulu

-0.60

 Pradesh

-0.59

SIZE

-0.59

 cumbers

-0.58

POSITIVE LOGITS

 extraord

0.99

lein

0.90

geist

0.88

iffe

0.86

lain

0.85

jee

0.84

bilt

0.84

="#

0.82

sonian

0.81

ufact

0.81

Activations Density 0.055%

mentions of specific names, likely related to a news story or event

proper nouns related to people, likely focusing on names and roles

No Comments

No Known Activations