INDEX

Explanations

concluding sentences or summaries in formal or professional documents. The neuron tends to activate strongly for phrases that wrap up or synthesize information, particularly in contexts like reports, manuals, or academic papers.

oai_token-act-pair · claude-3-5-sonnet-20240620 Triggered by @danielepoterti

terms that indicate skills and competencies related to storytelling and production

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

Follows lowercase letters

np_acts-logits-general · gemini-2.0-flash

religious and national concepts

np_acts-logits-general · gemini-2.5-flash-lite

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GEMMA-2-2B @ 24-gemmascope-res-16k

Configuration

google/gemma-scope-2b-pt-res/layer_24/width_16k/average_l0_73

Prompts (Dashboard)

36,864 prompts, 128 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Features

16,384

Data Type

float32

Hook Name

blocks.24.hook_resid_post

Hook Layer

Architecture

jumprelu

Context Size

1,024

Dataset

monology/pile-uncopyrighted

Activation Function

relu

Embeds

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

 awfully

-0.82

 wherewith

-0.78

 heretofore

-0.77

tvguidetime

-0.77

 unknownFields

-0.74

 terribly

-0.68

 totally

-0.68

 whence

-0.67

 appreciably

-0.66

 geni

-0.65

POSITIVE LOGITS

 courants

1.09

 preuves

1.05

 démocr

1.00

 prisonniers

0.98

 sauvages

0.94

 nationaux

0.94

 religieuses

0.94

 définiti

0.94

 prêtres

0.93

 chré

0.92

Activations Density 0.028%

concluding sentences or summaries in formal or professional documents. The neuron tends to activate strongly for phrases that wrap up or synthesize information, particularly in contexts like reports, manuals, or academic papers.

terms that indicate skills and competencies related to storytelling and production

Follows lowercase letters

religious and national concepts

No Comments

No Known Activations