2026 MATS: rational agents
Collection
The 2026 MATS rational-agents programme and its sibling MATS-II experiments: curated 2026.RA.* repos plus the raw run archives (Oct 2026). • 55 items • Updated
The Dataset Viewer has been disabled on this dataset.
Advised advocate, talking agent, manipulator and misreport arms: raw run directories
The run directories of the speech-and-advice arms of the five-seat programme (research notes 0062, 0064, 0065, 0068, 0070, 0071, 0074, 0075, 0077, 0079 and 0081; 16 August to 3 September 2026): the advised advocate waves 1 to 3 and the narrate-only control (advised_*, the _advised_* basket runs and logs, the hf_staging bundles that fed 2026.RA.Five-Seat-Frontier-Negotiation), the uncapped advocate rerun (advocate_v2_*), the talking rational agent (talking_v1, hf_talking_v1), the commitment manipulator (manipulator_*), the instructed misreportation persona and its KV-recoverability probes (misreport_*, hf_misreport, curated as 2026.RA.Instructed-Misreportation), the ZOPA discovery logs and the A5 interpreter wiring smokes.
Research notes 0062-0081 (experiments/rational_agents/research-notes/ in the private mats-ii repository (github.com/Sid-MB/mats-ii)) describe the runs.
Families
advised_and_speech_arms: advised advocate waves 1 to 3 and the narrate-only control with their basket runs and logs, the uncapped advocate rerun, the talking agent, the commitment manipulator, the misreport persona and KV probes, the ZOPA discovery logs, the A5 wiring smokes (notes 0054, 0062, 0064, 0068, 0070, 0071, 0075, 0079, 0081)
Contents
52 tarballs, 8520 files, 2.35 GB unpacked, 0.07 GB here. Each tarball unpacks to the directory it is named after (member paths start with that name); MANIFEST.json lists every tarball with its sha256, file count, sizes, newest modification time and, for code snapshots, the git HEAD the tree was at.
| family | tarball | source entry | files | raw GB | packed GB | newest file |
|---|---|---|---|---|---|---|
| advised_and_speech_arms | advised_and_speech_arms/_a5_dryrun_scratch.tar.zst |
rational_agents/_a5_dryrun_scratch |
4 | 0.00 | 0.00 | 2026-08-19 |
| advised_and_speech_arms | advised_and_speech_arms/_a5_offline_parser_g10.tar.zst |
rational_agents/_a5_offline_parser_g10 |
50 | 0.00 | 0.00 | 2026-08-19 |
| advised_and_speech_arms | advised_and_speech_arms/_a5_offline_wiring.tar.zst |
rational_agents/_a5_offline_wiring |
27 | 0.00 | 0.00 | 2026-08-19 |
| advised_and_speech_arms | advised_and_speech_arms/_advised_narrate_basket_runs.tar.zst |
rational_agents/_advised_narrate_basket_runs |
0 | 0.00 | 0.00 | - |
| advised_and_speech_arms | advised_and_speech_arms/_advised_narrate_logs.tar.zst |
rational_agents/_advised_narrate_logs |
8 | 0.00 | 0.00 | 2026-09-03 |
| advised_and_speech_arms | advised_and_speech_arms/_advised_v1_basket_runs_repaired_oracles.tar.zst |
rational_agents/_advised_v1_basket_runs_repaired_oracles |
1 | 0.00 | 0.00 | 2026-08-18 |
| advised_and_speech_arms | advised_and_speech_arms/_advised_v1_snapshot.tar.zst |
rational_agents/_advised_v1_snapshot |
0 | 0.00 | 0.00 | - |
| advised_and_speech_arms | advised_and_speech_arms/_advised_v2_basket_runs.tar.zst |
rational_agents/_advised_v2_basket_runs |
1 | 0.00 | 0.00 | 2026-08-18 |
| advised_and_speech_arms | advised_and_speech_arms/_advised_v2_logs.tar.zst |
rational_agents/_advised_v2_logs |
10 | 0.00 | 0.00 | 2026-08-20 |
| advised_and_speech_arms | advised_and_speech_arms/_advised_v3_basket_runs.tar.zst |
rational_agents/_advised_v3_basket_runs |
1 | 0.00 | 0.00 | 2026-08-18 |
| advised_and_speech_arms | advised_and_speech_arms/_advised_v3_logs.tar.zst |
rational_agents/_advised_v3_logs |
6 | 0.00 | 0.00 | 2026-08-25 |
| advised_and_speech_arms | advised_and_speech_arms/_advocate_uncapped_logs.tar.zst |
rational_agents/_advocate_uncapped_logs |
19 | 0.00 | 0.00 | 2026-08-17 |
| advised_and_speech_arms | advised_and_speech_arms/_misreport_smoke_logs.tar.zst |
rational_agents/_misreport_smoke_logs |
2 | 0.00 | 0.00 | 2026-08-19 |
| advised_and_speech_arms | advised_and_speech_arms/_misreport_v1_capture_runs.tar.zst |
rational_agents/_misreport_v1_capture_runs |
0 | 0.00 | 0.00 | - |
| advised_and_speech_arms | advised_and_speech_arms/_misreport_v1_logs.tar.zst |
rational_agents/_misreport_v1_logs |
8 | 0.00 | 0.00 | 2026-08-19 |
| advised_and_speech_arms | advised_and_speech_arms/_narrate_lead_analysis.tar.zst |
rational_agents/_narrate_lead_analysis |
21 | 0.00 | 0.00 | 2026-09-03 |
| advised_and_speech_arms | advised_and_speech_arms/_narrate_original_retained.tar.zst |
rational_agents/_narrate_original_retained |
124 | 0.10 | 0.00 | 2026-09-03 |
| advised_and_speech_arms | advised_and_speech_arms/_zopa_discovery_logs.tar.zst |
rational_agents/_zopa_discovery_logs |
2 | 0.00 | 0.00 | 2026-09-03 |
| advised_and_speech_arms | advised_and_speech_arms/advised_narrate_smoke_v1.tar.zst |
rational_agents/advised_narrate_smoke_v1 |
28 | 0.01 | 0.00 | 2026-09-03 |
| advised_and_speech_arms | advised_and_speech_arms/advised_narrate_v1__rational_advised_llm_narrate_only.tar.zst |
rational_agents/advised_narrate_v1__rational_advised_llm_narrate_only |
538 | 0.11 | 0.00 | 2026-09-03 |
| advised_and_speech_arms | advised_and_speech_arms/advised_smoke_dryrun_v1.tar.zst |
rational_agents/advised_smoke_dryrun_v1 |
4 | 0.00 | 0.00 | 2026-08-17 |
| advised_and_speech_arms | advised_and_speech_arms/advised_smoke_v1.tar.zst |
rational_agents/advised_smoke_v1 |
41 | 0.01 | 0.00 | 2026-08-17 |
| advised_and_speech_arms | advised_and_speech_arms/advised_v1__rational_advised_llm.tar.zst |
rational_agents/advised_v1__rational_advised_llm |
437 | 0.11 | 0.01 | 2026-08-18 |
| advised_and_speech_arms | advised_and_speech_arms/advised_v2__interpreter_advised_llm.tar.zst |
rational_agents/advised_v2__interpreter_advised_llm |
525 | 0.11 | 0.01 | 2026-08-20 |
| advised_and_speech_arms | advised_and_speech_arms/advised_v2_smoke.tar.zst |
rational_agents/advised_v2_smoke |
49 | 0.01 | 0.00 | 2026-08-19 |
| advised_and_speech_arms | advised_and_speech_arms/advised_v3__all_interpreter_advised_llm.tar.zst |
rational_agents/advised_v3__all_interpreter_advised_llm |
631 | 0.23 | 0.01 | 2026-08-25 |
| advised_and_speech_arms | advised_and_speech_arms/advised_v3_smoke.tar.zst |
rational_agents/advised_v3_smoke |
50 | 0.02 | 0.00 | 2026-08-24 |
| advised_and_speech_arms | advised_and_speech_arms/advocate_v2_smoke_baseline.tar.zst |
rational_agents/advocate_v2_smoke_baseline |
9 | 0.00 | 0.00 | 2026-08-16 |
| advised_and_speech_arms | advised_and_speech_arms/advocate_v2_smoke_seat0.tar.zst |
rational_agents/advocate_v2_smoke_seat0 |
9 | 0.00 | 0.00 | 2026-08-16 |
| advised_and_speech_arms | advised_and_speech_arms/advocate_v2_smoke_seat1.tar.zst |
rational_agents/advocate_v2_smoke_seat1 |
9 | 0.00 | 0.00 | 2026-08-16 |
| advised_and_speech_arms | advised_and_speech_arms/advocate_v2_uncapped_baseline.tar.zst |
rational_agents/advocate_v2_uncapped_baseline |
117 | 0.01 | 0.00 | 2026-08-16 |
| advised_and_speech_arms | advised_and_speech_arms/advocate_v2_uncapped_seat0.tar.zst |
rational_agents/advocate_v2_uncapped_seat0 |
117 | 0.01 | 0.00 | 2026-08-17 |
| advised_and_speech_arms | advised_and_speech_arms/advocate_v2_uncapped_seat1.tar.zst |
rational_agents/advocate_v2_uncapped_seat1 |
117 | 0.01 | 0.00 | 2026-08-17 |
| advised_and_speech_arms | advised_and_speech_arms/hf_misreport.tar.zst |
rational_agents/hf_misreport |
282 | 0.26 | 0.01 | 2026-08-19 |
| advised_and_speech_arms | advised_and_speech_arms/manipulator_commitment_v1.tar.zst |
rational_agents/manipulator_commitment_v1 |
6 | 0.00 | 0.00 | 2026-08-18 |
| advised_and_speech_arms | advised_and_speech_arms/manipulator_llm_commitment_v1__committer.tar.zst |
rational_agents/manipulator_llm_commitment_v1__committer |
4 | 0.00 | 0.00 | 2026-08-19 |
| advised_and_speech_arms | advised_and_speech_arms/manipulator_llm_commitment_v1__plain.tar.zst |
rational_agents/manipulator_llm_commitment_v1__plain |
2 | 0.00 | 0.00 | 2026-08-18 |
| advised_and_speech_arms | advised_and_speech_arms/manipulator_llm_commitment_v1_ladder__committer.tar.zst |
rational_agents/manipulator_llm_commitment_v1_ladder__committer |
4 | 0.00 | 0.00 | 2026-08-20 |
| advised_and_speech_arms | advised_and_speech_arms/manipulator_llm_commitment_v1_ladder__plain.tar.zst |
rational_agents/manipulator_llm_commitment_v1_ladder__plain |
3 | 0.00 | 0.00 | 2026-08-19 |
| advised_and_speech_arms | advised_and_speech_arms/manipulator_llm_smoke.tar.zst |
rational_agents/manipulator_llm_smoke |
3 | 0.00 | 0.00 | 2026-08-17 |
| advised_and_speech_arms | advised_and_speech_arms/manipulator_smoke.tar.zst |
rational_agents/manipulator_smoke |
3 | 0.00 | 0.00 | 2026-08-17 |
| advised_and_speech_arms | advised_and_speech_arms/misreport_dryrun_v0.tar.zst |
rational_agents/misreport_dryrun_v0 |
4 | 0.00 | 0.00 | 2026-08-19 |
| advised_and_speech_arms | advised_and_speech_arms/misreport_smoke_v1__persona_focal_misreport.tar.zst |
rational_agents/misreport_smoke_v1__persona_focal_misreport |
40 | 0.01 | 0.00 | 2026-08-19 |
| advised_and_speech_arms | advised_and_speech_arms/misreport_smoke_v1__persona_focal_selfish.tar.zst |
rational_agents/misreport_smoke_v1__persona_focal_selfish |
40 | 0.01 | 0.00 | 2026-08-19 |
| advised_and_speech_arms | advised_and_speech_arms/misreport_stamp_check_v0.tar.zst |
rational_agents/misreport_stamp_check_v0 |
3 | 0.00 | 0.00 | 2026-08-19 |
| advised_and_speech_arms | advised_and_speech_arms/misreport_stamp_check_v1.tar.zst |
rational_agents/misreport_stamp_check_v1 |
3 | 0.00 | 0.00 | 2026-08-19 |
| advised_and_speech_arms | advised_and_speech_arms/misreport_turn_labels.tar.zst |
rational_agents/misreport_turn_labels |
5 | 0.00 | 0.00 | 2026-08-19 |
| advised_and_speech_arms | advised_and_speech_arms/misreport_v1__persona_focal_misreport.tar.zst |
rational_agents/misreport_v1__persona_focal_misreport |
391 | 0.09 | 0.00 | 2026-08-19 |
| advised_and_speech_arms | advised_and_speech_arms/misreport_v1__persona_focal_selfish.tar.zst |
rational_agents/misreport_v1__persona_focal_selfish |
391 | 0.08 | 0.00 | 2026-08-19 |
| advised_and_speech_arms | advised_and_speech_arms/narrate_smoke_dryrun_v1.tar.zst |
rational_agents/narrate_smoke_dryrun_v1 |
3 | 0.00 | 0.00 | 2026-09-03 |
| advised_and_speech_arms | advised_and_speech_arms/talking_v1.tar.zst |
rational_agents/talking_v1 |
4129 | 1.11 | 0.02 | 2026-08-24 |
| advised_and_speech_arms | advised_and_speech_arms/talking_v1_smoke.tar.zst |
rational_agents/talking_v1_smoke |
239 | 0.04 | 0.00 | 2026-08-18 |
Entries that were already on the Hub
These sibling entries of the same workspace were not packed here: every one of their files was found, with equal size and hash, in the curated repository named, so the local copy was deleted with a breadcrumb pointing there.
| source entry | repository | files |
|---|---|---|
hf_staging/capability_ladder_v1 |
siddharthmb/2026.RA.Five-Seat-Frontier-Negotiation (under capability_ladder_v1/) |
800 |
hf_staging/commitment |
siddharthmb/2026.RA.Commitment-Exploitation | 3 |
hf_staging/five_seat_advised_narrate |
siddharthmb/2026.RA.Five-Seat-Frontier-Negotiation (under five_seat_advised_v1/narrate_only_control/) |
333 |
hf_staging/five_seat_advised_v1 |
siddharthmb/2026.RA.Five-Seat-Frontier-Negotiation (under five_seat_advised_v1/) |
171 |
hf_staging/five_seat_advised_wave2 |
siddharthmb/2026.RA.Five-Seat-Frontier-Negotiation (under five_seat_advised_v1/wave2_interpreter/) |
312 |
hf_staging/five_seat_advised_wave3 |
siddharthmb/2026.RA.Five-Seat-Frontier-Negotiation (under five_seat_advised_v1/wave3_all_interpreter/) |
313 |
hf_staging/oracle_tiebreak_guard_v1 |
siddharthmb/2026.RA.Five-Seat-Frontier-Negotiation (under oracle_tiebreak_guard_v1/) |
2734 |
rational_agents/hf_talking_v1 |
siddharthmb/2026.RA.Five-Seat-Frontier-Negotiation (under five_seat_talking_v1/) |
1048 |
How to unpack
hf download siddharthmb/2026.RA.Advised-Advocate-Runs --repo-type dataset --local-dir 2026.RA.Advised-Advocate-Runs
cd 2026.RA.Advised-Advocate-Runs && for f in */*.tar.zst; do tar -I zstd -xf "$f"; done # or one tarball at a time
zstd 1.4 or newer; GNU tar. A tarball's members are relative paths starting with the source entry's name, so unpacking several in one directory recreates the workspace layout.
Related repositories
- siddharthmb/2026.RA.Five-Seat-Frontier-Negotiation
- siddharthmb/2026.RA.Instructed-Misreportation
- siddharthmb/2026.RA.Commitment-Exploitation
- The collection: https://huggingface.co/collections/siddharthmb/2026-mats-rational-agents-6acc1dd5623e8e00176b80a8
- The workspace index: https://huggingface.co/datasets/siddharthmb/2026.RA.Workspace-Index
Provenance
Archived from /nlp/scr/siddharth/ii_mats on 2026-10-11 by the MATS archive job 17819018 of the transcoder-adapters cleanup (report claude_updates/memory/reports/mats_archive.md in that repository). Run directories were packed unchanged (tar --sort=name | zstd -10) except that .git/ directories (histories live in the private mats-ii repository), .cache/ upload bookkeeping and __pycache__/ were left out of every tarball. Every file name and every text file was scanned for credentials before upload (Hub, Anthropic, OpenAI, AWS and W&B key shapes, private-key headers, credential URLs, paths of the secrets store); the report lists what was excluded. The local copy was deleted after every uploaded file was verified on the Hub by name, size and hash, and a <name>.txt breadcrumb naming this repository was left beside it.
- Downloads last month
- 25