INDEX

Explanations

references to sports players, contracts, and teams

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_fs3072-jb

Configuration

jbloom/GPT2-Small-Feature-Splitting-Experiment-Layer-8/blocks.8.hook_resid_pre_3072

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

Skylion007/openwebtext

Features

3,072

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

ories

-0.85

Operation

-0.81

ĺħ

-0.74

¿½

-0.73

utterstock

-0.70

ħĭ

-0.69

 Cooking

-0.69

½

-0.67

ouk

-0.66

een

-0.64

POSITIVE LOGITS

players

0.82

 drafted

0.79

acters

0.78

player

0.77

 endors

0.77

Unknown

0.77

 rosters

0.76

 corps

0.76

 kneeling

0.75

unknown

0.75

Activations Density 0.509%

references to sports players, contracts, and teams

No Comments

No Known Activations