INDEX

Explanations

phrases indicating perspective or recommendation

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

 nets

-0.63

calling

-0.62

dro

-0.60

pict

-0.59

 lines

-0.56

pie

-0.55

stores

-0.55

 rolled

-0.55

 netted

-0.55

 costing

-0.54

POSITIVE LOGITS

ilet

1.52

 summarize

1.15

 reiterate

1.13

 compensate

1.10

 illustrate

1.09

pping

1.09

ppings

1.07

 commemorate

1.06

asted

1.05

 clarify

1.05

Activations Density 0.527%

phrases indicating perspective or recommendation

No Comments

No Known Activations