INDEX

Explanations

structured numeric data or references related to values and statistics

oai_token-act-pair · gpt-4o-mini Triggered by @bot

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 8-res_fs49152-jb

Configuration

jbloom/GPT2-Small-Feature-Splitting-Experiment-Layer-8/blocks.8.hook_resid_pre_49152

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

Skylion007/openwebtext

Features

49,152

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

 disproportion

-0.72

 Bethlehem

-0.67

 delegation

-0.67

bed

-0.67

 dividing

-0.67

 braces

-0.64

cloth

-0.64

 blockers

-0.64

rons

-0.64

 Mellon

-0.63

POSITIVE LOGITS

~~~~~~~~

1.20

________________

1.18

~~~~~~~~~~~~~~~~

1.16

---------------

1.16

________________________________________________________________

1.15

-------------

1.10

--------------

1.10

------------------------

1.08

------------

1.05

****************

1.04

Activations Density 0.002%

structured numeric data or references related to values and statistics

No Comments

No Known Activations