INDEX

Explanations

words related to technical skills or training

oai_token-act-pair · gpt-3.5-turbo

phrases related to roles and relationships in various contexts

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 11-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.11.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.11.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

almost

-0.75

completely

-0.74

albeit

-0.73

Hopefully

-0.70

whatever

-0.70

umerable

-0.68

 EVEN

-0.68

Quite

-0.68

Very

-0.67

 fortunately

-0.67

POSITIVE LOGITS

 computers

0.90

 airplanes

0.90

 automobiles

0.89

 animals

0.88

 oneself

0.84

 toilets

0.84

 cars

0.84

 limbs

0.83

 bathrooms

0.83

 pets

0.82

Activations Density 0.498%

words related to technical skills or training

phrases related to roles and relationships in various contexts

No Comments

No Known Activations