INDEX

Explanations

references to locations characterized as "hotspots."

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

ŃĶ

-0.74

 syll

-0.71

yss

-0.71

 Rite

-0.69

©¶æ

-0.67

 restraint

-0.67

 guided

-0.66

 guidance

-0.65

ä»£

-0.64

 separation

-0.63

POSITIVE LOGITS

pots

1.72

pot

1.47

 hots

1.17

hots

1.06

peak

0.98

combe

0.96

hot

0.94

 Hots

0.93

stakes

0.93

pur

0.90

Activations Density 0.001%

references to locations characterized as "hotspots."

No Comments

No Known Activations