INDEX

Explanations

verbs related to actions and responsibilities

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

chrom

-0.79

ellen

-0.73

reddits

-0.72

 Seym

-0.70

Pie

-0.66

earch

-0.65

Hill

-0.64

 Skydragon

-0.64

ederation

-0.63

amine

-0.62

POSITIVE LOGITS

forward

0.87

 loads

0.84

 forward

0.82

weight

0.81

 weight

0.78

 onward

0.78

 carrier

0.74

 borne

0.73

loads

0.71

IGH

0.71

Activations Density 0.539%

verbs related to actions and responsibilities

No Comments

No Known Activations