INDEX

Explanations

words related to media events or press conferences

oai_token-act-pair · gpt-3.5-turbo

references to news and press conferences

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 11-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.11.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.11.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

netflix

-0.77

profits

-0.77

hemy

-0.70

 Enemies

-0.68

ukes

-0.67

vae

-0.65

azes

-0.65

inances

-0.64

ovies

-0.64

 Nebula

-0.63

POSITIVE LOGITS

 conference

1.54

 briefing

1.35

conference

1.25

 session

1.09

 availability

0.99

 seminar

0.98

 timeout

0.96

 appearance

0.96

 event

0.95

 ceremony

0.94

Activations Density 0.064%

words related to media events or press conferences

references to news and press conferences

No Comments

No Known Activations