FinQ · Qwen3.5-4B-Base + LoRA + decision head · 200k QA

Model naming and experiment labels

FinQ — Qwen3.5-4B-Base + LoRA + learned decision head, 200k distinct QA names this selected best checkpoint. FinL is the parallel full English-base six-round encoder route, and FinL-mlx is its MLX FP16 runtime. The original upstream model remains Qwen/Qwen3.5-4B-Base. Head-only, 50k and 9B Qwen checkpoints retain their explicit comparator names.

Paper: S²-MAS. The fixed-budget adaptation results and matched downstream methods define the comparison reported here.

Project background

S²-MAS is a society-scale multi-agent simulation framework for studying how household decisions affect consumption, cash reserves and payments. The paper connects a constructed resource of 500 million household-persona records, learned decision policies, reusable financial scenarios and a state-transition engine. Its purchase-funding application evaluates 12 million distinct synthetic household agents under paired conditions. Full-data FinL supplies the policy for that application; this repository supplies a fixed-budget adaptation checkpoint for the common decision-model comparison. Decision learning uses real-survey data. The GENESE/MAS engine implementation remains private.

The benchmark groups twelve systems into six decision-model routes: encoder option scoring; language models with learned decision heads; next-token decision SFT; dynamic-schema classification; multimodal encoder decisions; and provider decision interfaces. All receive text inputs in the common benchmark. The 17 model-repository entries comprise 10 fixed-budget trained variants—five native Laya and five Qwen variants—plus six checkpoint/runtime entries and one compatibility alias.

This model's role

FinQ follows the language-model plus learned decision-head route. LoRA adapters and a shared scalar decision head are trained together on Qwen3.5-4B-Base. The 200k head-only counterpart holds the backbone fixed, allowing the comparison to assess representation adaptation. The complete development partition selects this LoRA method before the separate 9B run.

Training design and checkpoint

• Method: lora.

• Distinct downstream training rows: 200,000 from the fixed decision-QA training partition.

• Initialization: Qwen/Qwen3.5-4B-Base.

• This stage terminal row exposures: 200,000; selected checkpoint stage exposures: 200,000.

• Selected checkpoint cumulative row exposures: 200,000; selected cumulative distinct rows: 200,000.

• A fixed 20,000-example development set selects this checkpoint. The complete 412,873-example development partition compares the two 200k Qwen4B adaptation methods.

Performance and matched comparisons

Language: English. Inference precision: BF16 Qwen backbone with FP32 candidate scoring.

The common held-out test contains 240,458 QA records, 119 question versions and 8,275 respondent/household groups. The checkpoint is selected using development data before this test.

Checkpoint / reference	Accuracy (%)	Question-macro F1 (%)	Invalid decisions
FinQ · Qwen3.5-4B-Base + LoRA + head · 200k QA	74.1435	36.7343	0
Qwen4B head only, same 200k unique QA	67.2841	30.8995	0

Accuracy weights every test record equally; invalid decisions count as incorrect. Question-macro F1 averages over every defined candidate class within each question version and then weights the 119 versions equally.

The head-only reference uses the same pretrained backbone, unique examples and common test, with its backbone fixed. This checkpoint also adapts the backbone representation through LoRA.

Package and inference

LoRA adapter plus a shared scalar decision head; obtain the official backbone separately.

The 4B initialization is the official pre-trained-only Base release.

• Model files: head.pt, adapter/adapter_config.json, adapter/adapter_model.safetensors.

• Reported common-test compute: BF16 Qwen backbone with FP32 candidate scoring.

• Obtain the official Qwen backbone and tokenizer named in the configuration separately.

• Inference scores the full state, question and complete candidate menu.

The supplied qwen_decision.py exposes the candidate-scoring inference interface. Supply the local package and the official backbone and tokenizer named in decision_config.json.

License and access

The project model contribution is offered under the S²-MAS Controlled Model Research License v1.2: noncommercial scientific use, no redistribution, and no unauthorized reverse engineering, extraction or distillation. Training and evaluation records are not distributed. Source-data access requires the provider's authorization. Permission requests: yli9919@hku.hk.

Project repository: s2-mas.