INDEX

Explanations

phrases related to news headlines or updates

oai_token-act-pair · gpt-3.5-turbo

references to significant entities or topics in news articles

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

SPONSORED

-0.80

qus

-0.77

 whilst

-0.72

 preceded

-0.72

table

-0.71

rats

-0.71

tics

-0.71

respect

-0.69

 Secondly

-0.69

 namely

-0.69

POSITIVE LOGITS

 Latest

1.10

 nation

1.06

 deadliest

1.05

 embattled

1.05

 largest

1.04

 latest

1.02

resa

0.99

 secretive

0.96

 National

0.96

 controversial

0.93

Activations Density 0.441%

phrases related to news headlines or updates

references to significant entities or topics in news articles

No Comments

No Known Activations