INDEX

Explanations

words relating to specific codes or names, potentially related to espionage, criminal activities, or investigations

oai_token-act-pair · gpt-3.5-turbo

references to codes or coded terminology

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 10-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.10.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.10.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

ties

-0.81

esville

-0.77

kus

-0.75

athi

-0.73

itor

-0.71

riks

-0.71

iatus

-0.70

 experien

-0.68

swer

-0.68

 Ples

-0.68

POSITIVE LOGITS

 snippet

0.96

velop

0.95

breaker

0.87

breakers

0.84

otle

0.81

 snippets

0.81

breaking

0.77

 codes

0.74

 reuse

0.74

Chicken

0.72

Activations Density 0.023%

words relating to specific codes or names, potentially related to espionage, criminal activities, or investigations

references to codes or coded terminology

No Comments

No Known Activations