INDEX

Explanations

specific sequences of characters like symbols or codes

oai_token-act-pair · gpt-3.5-turbo

negations or refusals in a conversation

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

 radar

-0.71

 Japanese

-0.68

 scattering

-0.66

enic

-0.65

 slumped

-0.63

 photography

-0.63

utterstock

-0.62

 cyan

-0.61

 Slug

-0.59

 Taiwanese

-0.59

POSITIVE LOGITS

must

0.84

should

0.84

need

0.83

certain

0.83

shall

0.82

Â¢

0.82

catentry

0.78

bec

0.78

needed

0.78

lest

0.77

Activations Density 0.361%

specific sequences of characters like symbols or codes

negations or refusals in a conversation

No Comments

No Known Activations