INDEX

Explanations

instances of the word "to," particularly in a repetitive context indicating a significant action or intention

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

New Auto-Interp

Top Features by Cosine Similarity

Comparing With GPT2-SMALL @ 10-res_post_32k-oai

Configuration

jbloom/GPT2-Small-OAI-v5-32k-resid-post-SAEs/v5_32k_layer_10.pt

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

Skylion007/openwebtext

Features

32,768

Data Type

torch.float32

Hook Name

blocks.10.hook_resid_post

Hook Layer

Architecture

standard

Context Size

Dataset

Skylion007/openwebtext

Activation Function

topk

Embeds

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

 Palest

-0.68

 Timbers

-0.63

 continuing

-0.62

halla

-0.62

raltar

-0.61

auts

-0.60

boys

-0.60

Pru

-0.60

 forever

-0.59

 Ajax

-0.58

POSITIVE LOGITS

 resemble

0.90

 suspect

0.87

 worry

0.86

 sweat

0.82

 doubt

0.81

 wonder

0.81

 deterior

0.80

 realize

0.80

 unravel

0.79

 scratch

0.78

Activations Density 0.115%

instances of the word "to," particularly in a repetitive context indicating a significant action or intention

No Comments

No Known Activations