INDEX

Explanations

terms related to medical professions and procedures

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

hement

-0.70

plet

-0.64

 commits

-0.61

fou

-0.61

 latter

-0.61

 repro

-0.60

 taps

-0.59

 rolls

-0.59

qv

-0.58

HOU

-0.58

POSITIVE LOGITS

ccording

1.23

Spoiler

0.83

 Originally

0.83

 Initially

0.82

 Firstly

0.81

 Suppose

0.77

neys

0.76

Sometimes

0.76

 Throughout

0.75

Throughout

0.74

Activations Density 2.310%

terms related to medical professions and procedures

No Comments

No Known Activations