INDEX

Explanations

terms associated with deception or dishonesty

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With LLAMA3.1-8B @ 26-llamascope-res-32k

Configuration

fnlp/Llama3_1-8B-Base-LXR-8x/Llama3_1-8B-Base-L26R-8x

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

cerebras/SlimPajama-627B

Features

32,768

Data Type

bfloat16

Hook Name

blocks.26.hook_resid_post

Hook Layer

Architecture

jumprelu

Context Size

1,024

Dataset

cerebras/SlimPajama-627B

Activation Function

relu

Embeds

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

æĬ¼

-0.15

alnÄ±z

-0.14

Ø±ÙĪÙħ

-0.14

CHandle

-0.14

ungi

-0.14

áº¹p

-0.14

avor

-0.13

Ð´Ð¸Ð½Ð°

-0.13

umi

-0.13

opic

-0.13

POSITIVE LOGITS

lie

0.84

 lies

0.82

 lying

0.74

Lie

0.70

 Lies

0.70

Lie

0.65

lie

0.63

 lied

0.59

lies

0.55

 liar

0.55

Activations Density 0.487%

terms associated with deception or dishonesty

No Comments

No Known Activations