INDEX

Explanations

phrases indicating personal beliefs or opinions

oai_token-act-pair · gpt-3.5-turbo

statements of belief or conviction

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

PlotsExplanationShow Test FieldDefault Test Text

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

cloth

-0.75

insert

-0.70

conservancy

-0.70

mentioned

-0.68

practice

-0.67

yna

-0.67

aste

-0.66

effect

-0.64

umber

-0.62

nice

-0.61

POSITIVE LOGITS

orean

0.76

 believe

0.74

orea

0.73

ieve

0.73

 believes

0.70

ĳå£«

0.70

phas

0.70

rill

0.69

orians

0.68

POSE

0.66

Activations Density 0.041%

phrases indicating personal beliefs or opinions

statements of belief or conviction

No Comments

No Known Activations