INDEX

Explanations

phrases related to actions and consequences

oai_token-act-pair · gpt-3.5-turbo

references to significant events or decisions

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 8-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.8.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.8.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

acia

-0.58

rex

-0.56

Fax

-0.55

specified

-0.55

ename

-0.53

eton

-0.52

asse

-0.51

 procurement

-0.51

alf

-0.50

SQL

-0.50

POSITIVE LOGITS

 reminders

0.87

 remind

0.82

 thrill

0.80

 nostalgic

0.80

 nostalgia

0.80

 reminding

0.79

 reminis

0.78

 entertaining

0.77

 inspirational

0.77

 highlight

0.76

Activations Density 1.215%

phrases related to actions and consequences

references to significant events or decisions

No Comments

No Known Activations