INDEX

Explanations

phrases that imply conditions or terms of exchange

New Auto-Interp

Top Features by Cosine Similarity

Configuration

ckkissane/attn-saes-gpt2-small-all-layers/gpt2-small_L5_Hcat_z_lr1.20e-03_l11.00e+00_ds49152_bs4096_dc1.00e-06_rsanthropic_rie25000_nr4_v9.pt

Prompts (Dashboard)

36,864 prompts, 128 tokens each

Dataset (Dashboard)

Skylion007/openwebtext

Features

49,152

Data Type

float32

Hook Name

blocks.5.attn.hook_z

Hook Layer

Architecture

standard

Context Size

128

Dataset

Skylion007/openwebtext

Activation Function

relu

Embeds

IFrame

Link

Head Attr Weights

0:0.01

1:0.01

2:0.03

3:0.05

4:0.16

5:0.02

6:0.12

7:0.40

8:0.02

9:0.03

10:0.05

11:0.04

Negative Logits

ussen

-1.80

 trenches

-1.74

 aisle

-1.74

ahime

-1.73

Bench

-1.66

 darkened

-1.62

uyomi

-1.61

 circles

-1.54

 corridors

-1.53

 corners

-1.53

POSITIVE LOGITS

 assurances

2.17

 donations

1.72

 surrender

1.70

 allegiance

1.68

 supplying

1.67

 reimburse

1.65

enance

1.59

 withholding

1.58

 membership

1.56

 unconditional

1.56

Activations Density 0.001%

phrases that imply conditions or terms of exchange

No Known Activations