INDEX

Explanations

technical or procedural terms, such as "process" or "known as"

oai_token-act-pair · gpt-3.5-turbo

references to various processes or methods described in the text

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 5-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.5.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.5.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

arling

-0.86

vertisement

-0.75

osponsors

-0.74

agascar

-0.70

owe

-0.68

 endors

-0.68

 Allies

-0.67

ervatives

-0.67

olls

-0.66

itol

-0.66

POSITIVE LOGITS

 ACTIONS

0.90

 chore

0.80

 procedures

0.78

 motions

0.76

pmwiki

0.75

 procedure

0.73

QUIRE

0.72

 rigorous

0.72

 evaluating

0.68

 induction

0.68

Activations Density 0.695%

technical or procedural terms, such as "process" or "known as"

references to various processes or methods described in the text

No Comments

No Known Activations