INDEX

Explanations

the word "sor" and "gran" with varying activation values

oai_token-act-pair · gpt-3.5-turbo

references to different types of sorghum and familial connections

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

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

aments

-0.73

lights

-0.72

 Bohem

-0.71

yright

-0.71

 Intake

-0.70

Commercial

-0.70

 Hague

-0.70

 Aviation

-0.70

 Lamp

-0.69

 Inspector

-0.67

POSITIVE LOGITS

sor

2.85

 gran

1.27

 retri

1.11

kin

1.11

EFF

1.09

 sher

1.09

 infer

1.07

fam

1.06

 Tide

1.06

 berries

1.02

Activations Density 0.059%

the word "sor" and "gran" with varying activation values

references to different types of sorghum and familial connections

No Comments

No Known Activations