INDEX

Explanations

negations or contradictions in the text

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_fs24576-jb

Configuration

jbloom/GPT2-Small-Feature-Splitting-Experiment-Layer-8/blocks.8.hook_resid_pre_24576

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

Skylion007/openwebtext

Features

24,576

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

 spor

-0.72

tein

-0.64

 creations

-0.63

crit

-0.62

 propelled

-0.61

 rotated

-0.61

 indirectly

-0.60

que

-0.59

friends

-0.59

 facult

-0.59

POSITIVE LOGITS

 unanim

0.83

umerable

0.81

eway

0.79

enough

0.75

 shortage

0.75

omore

0.75

 enough

0.71

yet

0.71

 overlap

0.70

 anymore

0.69

Activations Density 0.055%

negations or contradictions in the text

No Comments

No Known Activations