INDEX

Explanations

expressions of planned or scheduled actions

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

 understands

-0.62

ById

-0.61

 indices

-0.61

 matters

-0.61

holes

-0.60

rain

-0.60

 currents

-0.59

 contexts

-0.59

 Says

-0.59

 diapers

-0.58

POSITIVE LOGITS

 planned

0.79

 hoped

0.77

lling

0.75

etheus

0.73

lled

0.71

{{

0.70

llah

0.69

atham

0.69

pload

0.67

dule

0.66

Activations Density 0.111%

expressions of planned or scheduled actions

No Comments

No Known Activations