salakash/SamKash-Tolstoy

Add “compression mode”

#49 opened 22 days ago by

jbakerx

Clean dialogue punctuation and attribution

#48 opened 22 days ago by

jbakerx

Add continuous evaluation + regression tests

#47 opened 22 days ago by

jbakerx

Cross-lingual style alignment with parallel passages

#46 opened 22 days ago by

jbakerx

Reproducibility packaging (big for arXiv credibility)

#45 opened 22 days ago by

salakash

Negative sampling: teach what not to do

#44 opened about 1 month ago by

jbakerx

Add structural tokens for literature

1

#43 opened about 1 month ago by

jbakerx

Hierarchical prompting: outline → draft → polish

#42 opened about 1 month ago by

jbakerx

“Style strength” knob via LoRA scaling

1

#41 opened about 1 month ago by

jbakerx

Adapter routing (Mixture of adapters)

#40 opened about 1 month ago by

jbakerx

Add explicit “avoid anachronisms” behavior

1

#39 opened about 2 months ago by

jbakerx

Curriculum learning by length

1

#38 opened about 2 months ago by

jbakerx

DAPT → SFT → DPO (3-step quality upgrade)

1

#37 opened about 2 months ago by

jbakerx

Preference data bootstrapping (cheap but effective)

1

#36 opened about 2 months ago by

jbakerx

Multi author literary adapters + author-mix control

1

#35 opened about 2 months ago by

jbakerx

Add a lightweight style classifier for filtering

1

#34 opened about 2 months ago by

jbakerx

Better segmentation: train on scenes, not entire books

1

#33 opened about 2 months ago by

jbakerx

Add explicit “avoid anachronisms” behavior

1

#32 opened about 2 months ago by

jbakerx

RAG for factual grounding (optional)

1

#31 opened about 2 months ago by

jbakerx

Quantization aware adapter testing

1

#30 opened about 2 months ago by

jbakerx

Add continuous evaluation + regression tests

1

#29 opened about 2 months ago by

jbakerx

Better segmentation: train on scenes, not entire books

#28 opened about 2 months ago by

jbakerx

Add a lightweight style classifier for filtering

1

#27 opened about 2 months ago by

jbakerx

Preference data bootstrapping (cheap but effective)

1

#26 opened about 2 months ago by

jbakerx

Multi-author literary adapters + author-mix control

#25 opened about 2 months ago by

jbakerx

DAPT → SFT → DPO (3 step quality upgrade)

1

#24 opened about 2 months ago by

jbakerx

Add anti-quote-memorization checks

1

#23 opened about 2 months ago by

jbakerx

Shipping/serving improvements

#22 opened about 2 months ago by

jbakerx

Add a “modern-to-classical rewrite” mode

1

#21 opened about 2 months ago by

jbakerx

Create multiple adapters by sub skill

#20 opened about 2 months ago by

jbakerx

Add “character + POV” conditioning

#19 opened about 2 months ago by

jbakerx

Improve long form coherence with “chapter packing”

1

#18 opened about 2 months ago by

jbakerx

Use a Russian-friendly tokenizer strategy

#17 opened about 2 months ago by

jbakerx

Improve dataset hygiene for literary corpora

#16 opened about 2 months ago by

jbakerx

Evaluation you can publish in a follow-up paper

#15 opened about 2 months ago by

jbakerx

Release two adapters instead of one

1

#14 opened about 2 months ago by

jbakerx

Dual-lingual style transfer (English ↔ Russian)

#13 opened about 2 months ago by

jbakerx

Add Russian-language capability (two-stage LoRA)

1

#12 opened about 2 months ago by

jbakerx

Improve dataset hygiene for literary corpora

#11 opened about 2 months ago by

jbakerx

🚩 Report: Spam

1

#9 opened about 2 months ago by

diimdeep

change "lora_bias" to 'true'

1

#7 opened 3 months ago by

jbakerx

Update README.md

#5 opened 3 months ago by

jbakerx

Languages spoken

1

#2 opened 3 months ago by

mirix

Update README.md (Remove Duplicates)

#1 opened 3 months ago by

qpqpqpqpqpqp