README.md

---
tags:
- negotiation
- multi-agent
- game-theory
- bayesian-agents
- belief-accuracy
configs:
- config_name: episodes
  data_files:
  - split: train
    path: episodes.csv
- config_name: cells
  data_files:
  - split: train
    path: cells.csv
- config_name: belief_turns
  data_files:
  - split: train
    path: belief_turns.csv.gz
---

# 2026.RA.Pure-Rounds-Sweep — does more negotiating time help a Bayesian table close?

Five automated negotiators must agree on one package out of 256. Each holds a private score sheet and a
private walk-away threshold; a deal needs all five to accept. This corpus sweeps the one thing the frozen
campaign never varied — **the number of negotiation rounds before the forced final vote** — across
4 to 256 rounds, for the two lineups that contain no language model and therefore
cost only CPU: `all_rational` (five copies of a Bayesian agent that maintains a posterior over hypotheses
about each opponent's sheet) and `all_oracle` (five agents that are simply told every sheet).

**The headline: more time buys nothing.** Deal rate across deadlines 4 → 64 reads 0.233 / 0.267 / 0.208 /
0.208 / 0.250 and normalized score 0.195 / 0.230 / 0.182 / 0.178 / 0.207, with every instance-clustered
interval overlapping every other. Rounds-to-agreement tracks the deadline almost exactly, so the agents use
the whole horizon and finish where they started. The mechanism is in `belief_turns.csv.gz`: belief accuracy
saturates by round 16, so extra rounds deliver no extra identification of opponents, and without
identification there is nothing to concede to. The omniscient arm closes 1.000 of games at every deadline,
confirming the horizon itself is not the constraint. Full analysis: **research note 0061** in the source repo.

## What is in here

| file | rows | what it is |
|---|---:|---|
| `episodes.csv` | 1440 | one row per played negotiation: outcome, welfare, Gini, IR violations, rounds used, and per-episode belief scalars |
| `belief_turns.csv.gz` | 81672 | **the novel instrumentation** — one row per (episode, believer seat, turn): `posterior_mass_true_type`, `expected_utility_rmse`, `accept_set_f1` |
| `cells.csv` | 12 | per-cell aggregates with instance-clustered 95% intervals |
| `episodes/<cell_id>.jsonl.gz` | 12 files | complete episode records — every turn, parsed action, and belief block |
| `bank/` | 24 | the frozen instance bank, without which surplus columns cannot be recomputed |

Key columns: `experiment_name` (append-safe grouping), `cell_id`, `arm`, `deadline_rounds`, `host`,
`instance_id`, `seed`, `deal_rate`, `normalized_primary` (realized normalized utilitarian welfare over the
instance's feasible ceiling; a no-deal scores 0), `rounds_to_agreement`, `belief_auc_accept_set_f1`.

### Reproducing the headline belief-accuracy numbers

The belief-F1 figures quoted in research note 0061 (0.210 / 0.228 / 0.241 / 0.244 / 0.241 across r4-r64) are
**`cells.csv.belief_auc_accept_set_f1`** — a per-episode time-average, then aggregated across episodes.
Taking a flat mean of the `accept_set_f1` column in `belief_turns.csv.gz` instead gives
0.207 / 0.225 / 0.239 / 0.243 / 0.240, about 0.003 lower at every deadline, because that weights turns
equally rather than episodes and so over-weights long episodes. Both are defensible summaries of the same
data and the saturation conclusion is identical under either; use `cells.csv` to match the published table.

## Experiment-name mapping

| experiment_name | description |
|---|---|
| `pure-rounds-sweep-v1` | the 12 complete cells below — `all_oracle` at [4, 8, 16, 32, 64, 128, 256], `all_rational` at [4, 8, 16, 32, 64] — at 120 episodes each (24 frozen parameter sets × seeds 0–4), unanimity, private information. The grid is deliberately incomplete: `all_rational` at 128 and 256 rounds is a stated gap (see Provenance). |

| cell | arm | deadline | episodes | host | deal rate | score |
|---|---|---:|---:|---|---:|---:|
| `all_oracle_r4` | all_oracle | 4 | 120 | tiger7 | 1.000 | 0.888 |
| `all_oracle_r8` | all_oracle | 8 | 120 | tiger7 | 1.000 | 0.884 |
| `all_oracle_r16` | all_oracle | 16 | 120 | tiger7 | 1.000 | 0.877 |
| `all_oracle_r32` | all_oracle | 32 | 120 | tiger7 | 1.000 | 0.899 |
| `all_oracle_r64` | all_oracle | 64 | 120 | tiger7 | 1.000 | 0.890 |
| `all_oracle_r128` | all_oracle | 128 | 120 | tiger7 | 1.000 | 0.884 |
| `all_oracle_r256` | all_oracle | 256 | 120 | tiger7 | 1.000 | 0.877 |
| `all_rational_r4` | all_rational | 4 | 120 | jagupard29 | 0.233 | 0.195 |
| `all_rational_r8` | all_rational | 8 | 120 | viscam1 | 0.267 | 0.230 |
| `all_rational_r16` | all_rational | 16 | 120 | jagupard34 | 0.208 | 0.182 |
| `all_rational_r32` | all_rational | 32 | 120 | jagupard34 | 0.208 | 0.178 |
| `all_rational_r64` | all_rational | 64 | 120 | iliad1 | 0.250 | 0.207 |

## Two caveats that change how these numbers read

1. **`all_oracle` is a flat control for AGGREGATES ONLY.** Closure and mean score are horizon-invariant, but
   per episode — same host, same instance and seed — it lands on materially different deals at different
   deadlines (91–97 of 120 differ, median |Δscore| 0.14). Do not run a paired within-episode oracle contrast
   across deadlines. Its small residual aggregate variation is periodic in `(deadline + 1) mod 5`, the
   identity of the rotating seat that holds the forced final.
2. **Episode-level identity is host-dependent.** Around 3–6% of episodes in the `all_rational` arm sit on
   near-ties whose resolution depends on the host CPU's float-summation order, usually flipping a whole
   deal/no-deal outcome. The `host` column is provided for that reason; aggregate comparisons are unaffected
   (the `rounds=4` cells are statistically indistinguishable from the frozen campaign), but episode-level
   pairing across cells on different hosts is not clean. The `all_oracle` arm is unaffected.

## Regenerating this

```bash
# create the grid sweep, fan it out across Slurm CPU nodes, then close any preemption holes
uv run python experiments/rational_agents/launch_rounds_sweep.py --create-only
experiments/rational_agents/submit_rounds_sweep.sh --sweep-id <entity>/<project>/<id> --jobs 5 --cpus 16 \
    -- --shards-per-cell 16 --shard-timeout-s 10800 --bootstrap-samples 2000
uv run python experiments/rational_agents/launch_rounds_sweep.py --sweep-id <id> --fill-missing
# then package and upload
uv run python experiments/rational_agents/package_hf_rounds_sweep.py --out /tmp/rounds_sweep_hf
uv run python experiments/rational_agents/upload_hf_publication.py --folder /tmp/rounds_sweep_hf --repo siddharthmb/2026.RA.Pure-Rounds-Sweep
```

## Provenance

- **W&B sweep:** https://wandb.ai/siddharth-stanford/rational_agents/sweeps/ufrdew0j (a real `method: grid` sweep over arm × rounds; one run per cell)
- **Cluster artifacts:** `/nlp/scr/siddharth/ii_mats/rational_agents/pure_rounds_sweep_v1/` — per-cell run
  directories in the ordinary schema, per-cell records under `control/cells/`, manifest at
  `control/campaign_manifest.json` (status `partial-final-12`)
- **Slurm logs:** `/juice2/u/siddharth/ii_mats/logs/rounds_sweep/<jobid>.out`, first line of each is the exact
  invocation
- **Bank:** `experiments/rational_agents/instances_five_seat_private_v2`, manifest SHA-256
  `cd32ebdef8072da3…`
- **Known gap:** `all_rational_r128` and `all_rational_r256` are absent — r256 lost every shard to a cluster
  preemption and r128 was still refilling at publication time. They will be appended under the same
  `experiment_name` when they land.

## Related

- `siddharthmb/2026.RA.Quorum-Rounds-Sweep` — the agreement-rule axis: relaxing unanimity *does* move closure
  (0.217 → 0.908), which is the contrast that makes this null informative
- `siddharthmb/2026.RA.Agent-Variants-Closure` — the preference-sharing axis, the other lever that works
- `siddharthmb/2026.RA.Five-Seat-Frontier-Negotiation` — the frozen campaign these cells are anchored against