INDEX

Explanations

terms related to collaboration and secret agreements, often involving unethical or illegal activities

oai_token-act-pair · gpt-3.5-turbo

terms related to collusion and complicity 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 @ 1-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.1.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.1.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

¥µ

-0.83

Hop

-0.77

 Balanced

-0.76

robat

-0.70

 Rite

-0.68

ISS

-0.67

mac

-0.67

ĻĤ

-0.67

zl

-0.66

stood

-0.66

POSITIVE LOGITS

 collusion

1.37

 complicity

0.92

 tampering

0.79

 complicit

0.78

 wrongdoing

0.74

udicrous

0.74

 cartel

0.73

 wink

0.73

 rigging

0.72

 unfocusedRange

0.71

Activations Density 0.009%

terms related to collaboration and secret agreements, often involving unethical or illegal activities

terms related to collusion and complicity in various contexts

No Comments

No Known Activations