INDEX

Explanations

words related to watching videos or full content, and sometimes specific details like trailers or interviews

oai_token-act-pair · gpt-3.5-turbo

references to trailers and videos

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

rang

-0.83

ylum

-0.75

©¶æ¥µ

-0.68

Ħ¢

-0.68

cffff

-0.67

âĵĺ

-0.67

aeda

-0.65

bably

-0.65

 fortun

-0.65

#$

-0.64

POSITIVE LOGITS

 Highlights

0.93

 highlights

0.89

 below

0.88

 excerpt

0.88

 slideshow

0.82

 VIDEOS

0.81

 Transcript

0.80

 HERE

0.80

 latest

0.78

LAT

0.77

Activations Density 0.118%

words related to watching videos or full content, and sometimes specific details like trailers or interviews

references to trailers and videos

No Comments

No Known Activations