Raifu Wars โ€” Riko narrator, B4-9B (LoRA)

The voice of Riko Ricochet, an AI VTuber who plays Raifu Wars live and talks through her own match. A reinforcement-learning policy (RaifuWars-RL-ActionScorer-Cards) makes her moves; this model says what she is doing and why, reacts to the other players, and chats between turns.

It is a LoRA for B4-9B, distilled from a 27B narrator. On held-out prompts it scores slightly above that teacher, and it runs about three times faster on one 16 GB-class card. The merged model is RaifuWars-Riko-B4-9B.

Files

file use
adapter_model.safetensors, adapter_config.json PEFT adapter (r 32, alpha 64, attention and MLP projections)
RaifuWars-Riko-B4-9B-LoRA-f16.gguf the same adapter for llama.cpp: llama-server -m B4-9B-Q8_0.gguf --lora RaifuWars-Riko-B4-9B-LoRA-f16.gguf

All numbers below were measured the second way: B4-9B Q8_0 with this f16 LoRA loaded at runtime, thinking off.

What it was trained on

Her prompts are the live game: whose turn it is, what just happened, the standings, her own options, and a short read of the board (how far she is from the next tier, her best shot, who is threatening her). Prompts come from real Raifu Wars matches played by her policy.

teacher   B5-27B + Riko's narrator LoRA, one line for each of 6,049 narrator prompts
filter    a rule checker (format, length, pronouns, invented facts, stock phrases, echoing the
          report back) and a 27B judge (faithful to the game, owns her moves, score >= 4/5);
          the same line kept at most 3 times
student   B4-9B, SFT on the 5,120 lines that passed (85%), loss on her line only
          2 epochs, lr 1e-4 cosine, LoRA r 32 / alpha 64, max length 3072

The training lines are the teacher's own words, unedited: the student learns the teacher's best lines and none of its failures. The prompts and lines are not released.

Numbers

838 held-out narrator prompts, two lines each (1,676 lines), seed 99. "Good" means the line passed the rule checker and the judge.

narrator good hard fail judge flag judge score
B4-9B 57.2% 16.0% 22.6% 3.63
B4-9B + an earlier LoRA 70.7% 4.5% 19.6% 3.69
B5-27B + narrator LoRA (the teacher) 89.7% 1.7% 6.0% 3.97
this LoRA 92.1% 0.9% 3.9% 3.98

What it fixed in the base 9B, counted over the same 1,676 lines:

problem B4-9B this LoRA
wrong pronouns (every character in the game is a girl) 88 0
echoing the game report back 56 0
stock phrases 43 1
number dumps 27 1
invented hits 21 6
over-long lines 80 3

Limitations

  • The judge is the same kind of model that filtered the training lines. The prompts are new to the model, but part of the gain is learning what this judge likes. It has also been watched in a live narrated match and held up, but that is one match.
  • Pronoun nouns slip through. In its first live match it called another player "a proper gentleman". The checker missed nouns like that until then; it catches them now.
  • It is built for one character and one prompt format. Outside Riko's narrator prompts it is B4-9B with a habit of short, in-character game commentary.
Downloads last month
58
GGUF
Model size
58.2M params
Architecture
qwen35
Hardware compatibility
Log In to add your hardware

16-bit

Inference Providers NEW
This model isn't deployed by any Inference Provider. ๐Ÿ™‹ Ask for provider support

Model tree for yotisstudios/RaifuWars-Riko-B4-9B-LoRA