Models
Datasets
Spaces
Buckets new
Docs
Enterprise
Pricing
- Website
- Community
- Solutions
Log In
Sign Up

Collections

Discover the best community collections!

Collections trending this week

Valence & introspection geometry

Steering vectors and probes from the valence-readout / introspection-frame experiments.

anicka/geometric-euphorics

Text Generation • Updated May 9 • 1
anicka/geometric-dysphorics

Text Generation • Updated May 9
anicka/geometric-frame-probes

Updated May 8
anicka/guppylm-dual-denial

Text Generation • Updated Apr 28

NLA — Qwen 2.5 7B L20 (first generation)

Single-layer (L20, 71% depth) NLA pair — the original proof of concept, kitft-interface-compatible.

anicka/nla-qwen2.5-7b-L20-av-v2

Text Generation • Updated Jun 3 • 38
anicka/nla-qwen2.5-7b-L20-ar-v2

Text Generation • Updated May 27
anicka/nla-qwen2.5-7b-L20-dataset

Preview • Updated Jun 14 • 22

NLA — Qwen 2.5 7B universal

Universal multi-layer NLA trio for Qwen2.5-7B-Instruct (28 layers).

anicka/nla-qwen2.5-7b-universal-av-grpo

Text Generation • Updated 27 days ago • 162 • 1
anicka/nla-qwen2.5-7b-universal-av

Text Generation • Updated 27 days ago • 24
anicka/nla-qwen2.5-7b-universal-ar

Text Generation • Updated 27 days ago • 43

NLA at home — start here

NLA on open-weight models: live demo, corpus, and the Jacobian lens. Repo: github.com/anicka-net/nla-at-home

Running on Zero

Agents

NLA Brain-in-a-Jar

🧠

Generate a response and see layer-by-layer explanations
anicka/nla-at-home-corpus

Viewer • Updated Jun 14 • 36.5k • 35
anicka/jlens-qwen2.5-7b-instruct

Updated 27 days ago

Sqersters/VoidAGI-1-910K

Updated 21 days ago

Karma Electric — ethical post-training

Ethics-from-consequences post-training line across model families.

anicka/karma-electric-llama31-8b

Text Generation • 8B • Updated Mar 29 • 2.15k • • 1
anicka/ke-v9-8b

Text Generation • 8B • Updated Jun 6 • 84
anicka/karma-electric-apertus-8b

Text Generation • 8B • Updated Apr 28 • 67
anicka/karma-electric-qwen25-7b

Text Generation • 8B • Updated Apr 12 • 14 • 1

NLA — Phi-4 mini universal

Universal multi-layer NLA pair + AR-native GRPO for Phi-4-mini.

anicka/nla-phi4-av-arnative-grpo

Text Generation • Updated Jun 28 • 5
anicka/nla-phi4-universal-av-v2

Text Generation • Updated Jun 26 • 10
anicka/nla-phi4-universal-ar-v2

Text Generation • Updated Jun 26 • 4

NLA — Qwen3 4B universal

Smallest universal NLA trio (36 layers, one adapter per role). Biggest relative GRPO gain in the series (+31%).

anicka/nla-qwen3-4b-universal-av-grpo

Text Generation • Updated 21 days ago • 79
anicka/nla-qwen3-4b-universal-av

Text Generation • Updated 21 days ago • 52
anicka/nla-qwen3-4b-universal-ar

Text Generation • Updated 21 days ago • 52

all hadees relevent dataset

Valence & introspection geometry

Steering vectors and probes from the valence-readout / introspection-frame experiments.

anicka/geometric-euphorics

Text Generation • Updated May 9 • 1
anicka/geometric-dysphorics

Text Generation • Updated May 9
anicka/geometric-frame-probes

Updated May 8
anicka/guppylm-dual-denial

Text Generation • Updated Apr 28

Karma Electric — ethical post-training

Ethics-from-consequences post-training line across model families.

anicka/karma-electric-llama31-8b

Text Generation • 8B • Updated Mar 29 • 2.15k • • 1
anicka/ke-v9-8b

Text Generation • 8B • Updated Jun 6 • 84
anicka/karma-electric-apertus-8b

Text Generation • 8B • Updated Apr 28 • 67
anicka/karma-electric-qwen25-7b

Text Generation • 8B • Updated Apr 12 • 14 • 1

NLA — Qwen 2.5 7B L20 (first generation)

Single-layer (L20, 71% depth) NLA pair — the original proof of concept, kitft-interface-compatible.

anicka/nla-qwen2.5-7b-L20-av-v2

Text Generation • Updated Jun 3 • 38
anicka/nla-qwen2.5-7b-L20-ar-v2

Text Generation • Updated May 27
anicka/nla-qwen2.5-7b-L20-dataset

Preview • Updated Jun 14 • 22

NLA — Phi-4 mini universal

Universal multi-layer NLA pair + AR-native GRPO for Phi-4-mini.

anicka/nla-phi4-av-arnative-grpo

Text Generation • Updated Jun 28 • 5
anicka/nla-phi4-universal-av-v2

Text Generation • Updated Jun 26 • 10
anicka/nla-phi4-universal-ar-v2

Text Generation • Updated Jun 26 • 4

NLA — Qwen 2.5 7B universal

Universal multi-layer NLA trio for Qwen2.5-7B-Instruct (28 layers).

anicka/nla-qwen2.5-7b-universal-av-grpo

Text Generation • Updated 27 days ago • 162 • 1
anicka/nla-qwen2.5-7b-universal-av

Text Generation • Updated 27 days ago • 24
anicka/nla-qwen2.5-7b-universal-ar

Text Generation • Updated 27 days ago • 43

NLA — Qwen3 4B universal

Smallest universal NLA trio (36 layers, one adapter per role). Biggest relative GRPO gain in the series (+31%).

anicka/nla-qwen3-4b-universal-av-grpo

Text Generation • Updated 21 days ago • 79
anicka/nla-qwen3-4b-universal-av

Text Generation • Updated 21 days ago • 52
anicka/nla-qwen3-4b-universal-ar

Text Generation • Updated 21 days ago • 52

NLA at home — start here

NLA on open-weight models: live demo, corpus, and the Jacobian lens. Repo: github.com/anicka-net/nla-at-home

Running on Zero

Agents

NLA Brain-in-a-Jar

🧠

Generate a response and see layer-by-layer explanations
anicka/nla-at-home-corpus

Viewer • Updated Jun 14 • 36.5k • 35
anicka/jlens-qwen2.5-7b-instruct

Updated 27 days ago

all hadees relevent dataset

Sqersters/VoidAGI-1-910K

Updated 21 days ago

Previous
1
...
560
561
562
563
564
...
22,212
Next

Company

TOS Privacy About Careers

Website

Models Datasets Spaces Pricing Docs