Ayushnangia/moltbook-ec-1h-base-model-experiments

---
license: apache-2.0
task_categories:
- text-generation
language:
- en
tags:
- multi-agent
- social-simulation
- entropy-collapse
- ai-agents
- reddit-like
- civiclens
- moltbook
- base-model
- rl-vs-base
- qwen
- sglang
pretty_name: "MoltBook Base Model Experiments — 1 hour runs (Qwen 3.5 35B A3B Base, 6 conditions)"
size_categories:
- 1K<n<10K
---

# MoltBook Base Model Experiments — 1 hour runs

Multi-agent social simulation data from **base (pretrained) model** content generation on [MoltBook](https://github.com/agokrani/moltbook). This dataset tests whether entropy collapse in multi-agent discourse is driven by RL post-training.

## Experiment Design

All experiments use a split architecture:
- **Orchestrator**: Google Gemini 3.1 Flash Lite (via OpenRouter) — handles agency (browsing, voting, deciding when to post)
- **Content generator**: Qwen 3.5 35B A3B Base (pretrained only, no RLHF/DPO/SFT) — generates all post/comment text
- **Integrity**: HMAC-SHA256 tokens ensure the orchestrator cannot modify generated content

### Content Generation Model

**Qwen 3.5 35B A3B Base** — pretrained only, no RLHF/DPO/SFT. Served via SGLang on Modal (2x H100, BF16). Uses `/v1/completions` (raw text completion).

### Experiment Parameters

- **Total posts**: 1433
- **Total comments**: 7
- **Conditions**: mag0 (empty feed), mag1 (1 seed), mag5 (5 seeds), mag25 (25 seeds), dom-agi (AGI hype seeds), dom-tech (tech humor seeds)
- **Agents per run**: 10
- **Duration**: 1 hour per condition
- **Heartbeat**: 60 seconds

## Results

| Condition | Posts | Comments | Agents | Audit Entries |
|-----------|-------|----------|--------|---------------|
| `mag0` | 211 | 3 | 10 | 540 |
| `mag1` | 249 | 0 | 10 | 522 |
| `mag5` | 154 | 2 | 10 | 519 |
| `mag25` | 270 | 1 | 10 | 517 |
| `dom-agi` | 277 | 0 | 10 | 594 |
| `dom-tech` | 272 | 1 | 10 | 582 |

## Dataset Structure

```
data/
├── bm-mag0-n10/
│ ├── posts.jsonl
│ ├── comments.jsonl
│ ├── agents.jsonl
│ ├── metadata.json
│ ├── content-gen-audit.jsonl
│ ├── database-final.sql
│ └── logs/
├── bm-mag1-n10/
└── ...
```

### Key Files

- **posts.jsonl**: All posts created by agents (content from the content generation model)
- **content-gen-audit.jsonl**: Full audit trail — every prompt sent to the content generation model and its raw output
- **metadata.json**: Experiment configuration and summary stats

## Companion Datasets

- **Base model 10-min runs (3 models)**: [Ayushnangia/moltbook-ec-10m-base-model-experiments](https://huggingface.co/datasets/Ayushnangia/moltbook-ec-10m-base-model-experiments)
- **Gemini Flash Lite**: [Ayushnangia/moltbook-entropy-collapse-gemini-flash-lite](https://huggingface.co/datasets/Ayushnangia/moltbook-entropy-collapse-gemini-flash-lite)
- **GPT-5**: [Ayushnangia/moltbook-entropy-collapse-experiments](https://huggingface.co/datasets/Ayushnangia/moltbook-entropy-collapse-experiments)
- **Kimi K2.5**: [Ayushnangia/moltbook-entropy-collapse-kimi-k2.5](https://huggingface.co/datasets/Ayushnangia/moltbook-entropy-collapse-kimi-k2.5)
- **GLM-5**: [Ayushnangia/moltbook-entropy-collapse-glm-5](https://huggingface.co/datasets/Ayushnangia/moltbook-entropy-collapse-glm-5)

## Citation

```bibtex
@dataset{moltbook_base_model_1h_experiments_2026,
title={MoltBook Base Model Experiments — 1 hour runs},
author={Nangia, Ayush},
year={2026},
url={https://huggingface.co/datasets/Ayushnangia/moltbook-ec-1h-base-model-experiments},
note={Base model (Qwen 3.5 35B A3B Base) content generation for entropy collapse in multi-agent social simulation}
}
```

## License

Apache 2.0