INDEX

Explanations

references to specific chunks of items or entities

oai_token-act-pair · gpt-3.5-turbo

terms related to bursts and chunks, particularly in the context of scientific phenomena

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 0-res-jb

Configuration

jbloom/GPT2-Small-SAEs-Reformatted/blocks.0.hook_resid_pre

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

Skylion007/openwebtext

Features

24,576

Data Type

torch.float32

Hook Point

blocks.0.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

 downgrade

-0.76

 haircut

-0.76

 fence

-0.73

fit

-0.70

 carpet

-0.69

 trend

-0.69

 handic

-0.69

 valuation

-0.68

POS

-0.68

uph

-0.68

POSITIVE LOGITS

 bursts

2.63

 chunks

2.33

 fragments

2.00

 snippets

1.98

 flashes

1.98

 pulses

1.75

 batches

1.75

 slices

1.73

 increments

1.68

 strands

1.68

Activations Density 0.042%

references to specific chunks of items or entities

terms related to bursts and chunks, particularly in the context of scientific phenomena

No Comments

No Known Activations