INDEX

Explanations

mentions of being impressed or impressed by something/someone

oai_token-act-pair · gpt-3.5-turbo

expressions of admiration or being impressed

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 1-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.1.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.1.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

turn

-0.69

access

-0.68

rama

-0.68

violence

-0.67

 extraction

-0.67

 diverted

-0.63

 abort

-0.61

leasing

-0.61

population

-0.61

build

-0.61

POSITIVE LOGITS

 impressed

1.15

 resemb

0.85

sson

0.76

 Wenger

0.74

 surprised

0.73

MENTS

0.71

acebook

0.69

 admire

0.69

oresc

0.69

GGGGGGGG

0.69

Activations Density 0.009%

mentions of being impressed or impressed by something/someone

expressions of admiration or being impressed

No Comments

No Known Activations