INDEX

Explanations

sections separated by the character "Ċ", possibly indicating different news articles or stories

oai_token-act-pair · gpt-3.5-turbo

punctuation and formatting irregularities in sentences and phrases

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 10-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.10.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.10.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

 utilizing

-0.68

advertisement

-0.64

 utilized

-0.62

illac

-0.61

 tremend

-0.61

 favors

-0.59

,)

-0.59

auga

-0.59

 condos

-0.58

ngth

-0.57

POSITIVE LOGITS

BBC

1.00

Analysis

0.94

BBC

0.82

Labour

0.77

Media

0.76

 (Â£

0.73

Image

0.72

 Nato

0.69

 Â£

0.64

 Nasa

0.64

Activations Density 0.104%

sections separated by the character "Ċ", possibly indicating different news articles or stories

punctuation and formatting irregularities in sentences and phrases

No Comments

No Known Activations