INDEX

Explanations

names, likely of people or places

oai_token-act-pair · gpt-3.5-turbo

occurrences of the substring "lo" in words

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

 tremend

-0.74

krit

-0.69

 indebted

-0.63

 glim

-0.61

 Bethesda

-0.60

 principals

-0.60

 agric

-0.59

wee

-0.56

 transient

-0.56

 curfew

-0.56

POSITIVE LOGITS

onne

0.92

odor

0.86

ohyd

0.85

ãĤ¼ãĤ¦ãĤ¹

0.84

ades

0.79

reau

0.77

anwhile

0.77

anne

0.74

ucci

0.73

ello

0.73

Activations Density 0.081%

names, likely of people or places

occurrences of the substring "lo" in words

No Comments

No Known Activations