INDEX

Explanations

references to the news network "CNN"

oai_token-act-pair · gpt-3.5-turbo

mentions of the CNN network

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 0-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.0.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.0.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

————————————————

-0.70

 mages

-0.70

spell

-0.68

ãĢĳ

-0.68

-[

-0.68

)</

-0.67

 defences

-0.67

pine

-0.67

 mage

-0.66

POSITIVE LOGITS

CNN

3.51

CNN

3.08

 MSNBC

2.10

 CNBC

2.03

 HuffPost

1.64

CBS

1.61

PBS

1.59

 Newsweek

1.58

NBC

1.58

 ESPN

1.56

Activations Density 0.018%

references to the news network "CNN"

mentions of the CNN network

No Comments

No Known Activations