INDEX

Explanations

news story headlines

oai_token-act-pair · gpt-3.5-turbo

references to news stories or 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

ãĥĺãĥ©

-0.67

 phyl

-0.66

ergic

-0.64

esar

-0.62

onement

-0.59

essage

-0.59

 respectively

-0.59

 consecut

-0.58

 garage

-0.58

iate

-0.58

POSITIVE LOGITS

telling

1.20

 Highlights

1.10

 highlights

1.00

 Transcript

0.97

 Details

0.93

 Summary

0.88

 Stories

0.87

line

0.83

 Continued

0.81

 Story

0.79

Activations Density 0.014%

news story headlines

references to news stories or articles

No Comments

No Known Activations