INDEX

Explanations

year numbers and author names in a formatted structure frequently seen in academic references

oai_token-act-pair · gpt-3.5-turbo

references to authors and publication details

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

endi

-0.63

itiz

-0.61

 thous

-0.61

today

-0.60

VIDIA

-0.60

liners

-0.59

terday

-0.59

liner

-0.58

priority

-0.58

theless

-0.58

POSITIVE LOGITS

et

1.06

 Handbook

1.05

eds

0.94

ed

0.91

0.86

"'

0.86

0.85

Jr

0.85

0.84

Activations Density 0.061%

year numbers and author names in a formatted structure frequently seen in academic references

references to authors and publication details

No Comments

No Known Activations