INDEX

Explanations

advice or commentary related to principles or decisions

oai_token-act-pair · gpt-3.5-turbo Triggered by @bot

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 8-res_fs12288-jb

Configuration

jbloom/GPT2-Small-Feature-Splitting-Experiment-Layer-8/blocks.8.hook_resid_pre_12288

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

Skylion007/openwebtext

Features

12,288

Data Type

torch.float32

Hook Point

blocks.8.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

 Erica

-0.62

famous

-0.59

Ns

-0.58

 Victims

-0.58

ummies

-0.58

';

-0.57

/,

-0.57

 mysteries

-0.57

Orig

-0.57

OWN

-0.55

POSITIVE LOGITS

 reliance

0.81

ealous

0.79

 prolonged

0.78

 weakening

0.76

 shouldn

0.74

 inaction

0.73

 glance

0.73

 entails

0.73

 disruption

0.72

 could

0.72

Activations Density 0.746%

advice or commentary related to principles or decisions

No Comments

No Known Activations