INDEX

Explanations

acronyms related to professions or political subjects

oai_token-act-pair · gpt-3.5-turbo

forms of the prefix "pro-" followed by various suffixes

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

 similarities

-0.72

 ponds

-0.70

jriwal

-0.68

 wells

-0.66

 Skydragon

-0.65

¥µ

-0.64

HF

-0.63

 nurs

-0.61

 nests

-0.61

fine

-0.60

POSITIVE LOGITS

essional

0.85

cess

0.85

incial

0.85

etary

0.81

hetic

0.78

atural

0.77

imity

0.75

CTR

0.75

upt

0.75

cia

0.74

Activations Density 0.060%

acronyms related to professions or political subjects

forms of the prefix "pro-" followed by various suffixes

No Comments

No Known Activations