INDEX

Explanations

words related to personnel or groups of people, such as cast members or staff

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

PlotsExplanationShow Test FieldDefault Test Text

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

 accur

-0.66

 Fountain

-0.64

 Bout

-0.63

 Shame

-0.61

piration

-0.60

Occ

-0.60

rupt

-0.59

 Campaign

-0.58

 signage

-0.58

ashtra

-0.57

POSITIVE LOGITS

members

1.33

 members

1.18

 comprised

1.12

member

1.11

 member

1.09

 consisting

1.08

 mates

1.07

 comprising

1.07

 assembled

1.05

rons

1.02

Activations Density 1.451%

words related to personnel or groups of people, such as cast members or staff

No Comments

No Known Activations