INDEX

Explanations

words related to negative judgments on things or actions

oai_token-act-pair · gpt-3.5-turbo

strong negative descriptors of people or actions

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 0-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.0.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.0.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

 navig

-0.72

izons

-0.71

 Guid

-0.70

Haw

-0.70

ipolar

-0.68

igraph

-0.68

 recharge

-0.68

 Guitar

-0.67

 Piano

-0.67

sson

-0.67

POSITIVE LOGITS

 disgusting

2.69

 horrific

2.36

 vile

2.33

 horrifying

2.25

 horrendous

2.20

 horrible

2.20

 despicable

2.18

 hideous

2.08

 heinous

2.06

 gruesome

2.04

Activations Density 0.032%

words related to negative judgments on things or actions

strong negative descriptors of people or actions

No Comments

No Known Activations