INDEX

Explanations

references to specific buildings, particularly those known as palaces

oai_token-act-pair · gpt-3.5-turbo

references to "Palace" 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 @ 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

err

-0.71

arette

-0.70

schild

-0.69

CAST

-0.69

acted

-0.67

rss

-0.67

::::::::

-0.66

enegger

-0.66

went

-0.65

CHAT

-0.65

POSITIVE LOGITS

 Palace

0.94

osaurs

0.79

ibur

0.77

 intrigue

0.75

 palace

0.74

gur

0.74

 maiden

0.72

 renovation

0.71

 Hotel

0.70

holders

0.70

Activations Density 0.012%

references to specific buildings, particularly those known as palaces

references to "Palace" in various contexts

No Comments

No Known Activations