INDEX

Explanations

negations or contrasts in statements

oai_token-act-pair · gpt-3.5-turbo

negations or the phrase "not" in various contexts

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 8-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.8.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.8.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

velt

-0.73

çļ

-0.67

 arsen

-0.67

çīĪ

-0.66

éĥ

-0.65

ngth

-0.64

 throats

-0.64

ugh

-0.61

pez

-0.61

WAY

-0.59

POSITIVE LOGITS

 necessarily

1.41

withstanding

1.03

icably

0.96

icable

0.94

 exactly

0.93

 remotely

0.93

 entirely

0.91

 uncommon

0.85

yet

0.85

 solely

0.82

Activations Density 0.186%

negations or contrasts in statements

negations or the phrase "not" in various contexts

No Comments

No Known Activations