INDEX

Explanations

actions or abilities described in a positive light

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_fs768-jb

Configuration

jbloom/GPT2-Small-Feature-Splitting-Experiment-Layer-8/blocks.8.hook_resid_pre_768

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

Skylion007/openwebtext

Features

768

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

 Niet

-0.71

 Azerb

-0.64

 Borders

-0.61

 Grail

-0.61

 Seym

-0.61

 Nare

-0.58

 Clarkson

-0.56

 Frie

-0.56

 Jagu

-0.55

Gaw

-0.55

POSITIVE LOGITS

][

0.92

0.76

redients

0.69

0.66

];

0.65

].

0.65

::

0.65

actionDate

0.64

lement

0.64

Activations Density 4.238%

actions or abilities described in a positive light

No Comments

No Known Activations