INDEX

Explanations

references to a location named "Ram" and related events, such as political figures and trials

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

 Morales

-0.86

ABLE

-0.64

ahead

-0.63

REDACTED

-0.59

IAL

-0.58

cffff

-0.58

LECT

-0.57

 indications

-0.57

UCT

-0.57

lements

-0.57

POSITIVE LOGITS

ming

1.48

mers

1.40

pton

1.29

pling

1.25

mer

1.22

urai

1.22

blers

1.22

borgh

1.16

oufl

1.13

pled

1.10

Activations Density 0.957%

references to a location named "Ram" and related events, such as political figures and trials

No Comments

No Known Activations