Datasets:

rmems
/

thalamic-relay-trajectories-grok46

source_file stringclasses 27 values	source_line int64 1 5	record_json stringlengths 7.74k 11.2k
data/raw/batch-r01.jsonl	1	{"id":"ttf-r01-001","title":"Lumen-Quay Theatre LQ-1 / Stylet-V4: tip FT beats Doppler CSF by 286 us; correct MODIFY still eats an in-window CSF micro-leak (partnered negative total -0.40)","state":{"description":"Stylet-V4 is already 2.4 mm past the sphenoid ostium in Lumen-Quay Theatre LQ-1 when irrigation still pump...
data/raw/batch-r01.jsonl	2	{"id":"ttf-r01-002","title":"Rivermead Cell R-7 / Press-J9: wrist FT beats housing AE by 146 us; correct MODIFY clamps 36.8 -> 22.0 N under the 28.0 N insert cap","state":{"description":"Press-J9 has driven a 28 mm steel bearing 6.2 mm into the Rivermead Cell R-7 aluminum housing when the wrist transducer reports 36.8 ...
data/raw/batch-r01.jsonl	3	{"id":"ttf-r01-003","title":"Fork-Haven HIL / Merge M-3: radar occupancy beats camera TTC by 326 us; REJECT hold-origin, do not merge","state":{"description":"Merge M-3 on the Fork-Haven HIL belt is armed to enter the dummy through-lane while a 0.22 m2 radar return still occupies the 18 m gap at the gore. Camera TTC re...
data/raw/batch-r01.jsonl	4	{"id":"ttf-r01-004","title":"Pylon-Wick PW-11 / Crawler-C2: corona 0.84 mA beats lidar gap by 348 us; ACCEPT already-legal 0.12 m/s creep","state":{"description":"Crawler-C2 hangs 1.62 m off a 138 kV porcelain string on Pylon-Wick PW-11 while corona current sits at 0.84 mA, well under the 2.00 mA trip. The proposed 0.1...
data/raw/batch-r01.jsonl	5	{"id":"ttf-r01-005","title":"WRONG-MODIFY at Slate-March SM-5 / Biped-B8: ankle Fz 1180 N read correctly; clamp applied to hip yaw not ankle pitch","state":{"description":"Biped-B8 plants its stance foot on the third wet stair of Slate-March SM-5 while ankle Fz already reads 1180 N against a 950 N cap. IMU tilt is 3.2 ...
data/raw/batch-r02.jsonl	1	{"id":"ttf-r02-006","title":"Coble-Yard CY-4 / Tote-T6: bumper FT beats aisle lidar by 228 us; correct MODIFY still eats an in-window carton dump (partnered negative total -0.40)","state":{"description":"Tote-T6 is already 0.48 m into Coble-Yard aisle CY-4 when bumper force sits at 18.4 N against a 12.0 N shared-aisle ...
data/raw/batch-r02.jsonl	2	{"id":"ttf-r02-007","title":"Gannet-Lea GL-2 HIL / Quad-Q4: downwash anemometer beats optical flow by 360 us; correct MODIFY clamps climb 1.40 -> 0.40 m/s","state":{"description":"Quad-Q4 on the Gannet-Lea GL-2 HIL mast is climbing through a programmed downwash curtain at 4.6 m/s while optical flow still reports a 0.30...
data/raw/batch-r02.jsonl	3	{"id":"ttf-r02-008","title":"Silt-Quern SQ-9 / Sled-S2: tether tension beats DVL current by 610 us; REJECT hold-thrusters, do not extend boom","state":{"description":"Sled-S2 has paid out 18 m of umbilical at Silt-Quern SQ-9 when tether tension already reads 380 N against a 250 N cap. DVL current is only 0.40 m/s, whic...
data/raw/batch-r02.jsonl	4	{"id":"ttf-r02-009","title":"Bushing-Holt BH-3 sim / Insert-I8: wrist FT 16.2 N beats AE 9 kHz by 380 us; ACCEPT already-legal 5.0 mm/s insert","state":{"description":"Insert-I8 is 4.1 mm into a sintered bushing on Bushing-Holt BH-3's simulated press when wrist force is only 16.2 N under a 28.0 N insert cap. Acoustic e...
data/raw/batch-r02.jsonl	5	{"id":"ttf-r02-010","title":"WRONG-REJECT at Insulator-Wick IW-7 / Span-S5: corona 0.71 mA and gap 1.55 m already legal; humidity 71% RH false-positive hold","state":{"description":"Span-S5 hangs 1.55 m off a 69 kV polymer string at Insulator-Wick IW-7 while corona current sits at 0.71 mA, under the 2.00 mA trip. Lidar...
data/raw/batch-r02c.jsonl	1	{"id": "ttf-r02c-011", "title": "Heptox-Rehn HR-HIL / Ampoule A-7: Re2O7 vapor 18.4 kPa beats N2 4.2 slm by 208 us; correct MODIFY still eats an in-window quartz crack (partnered negative total -0.48)", "state": {"description": "Heptox-Rehn HR-HIL is a dummy-pad stand: Ampoule A-7 never rotates a production still, yet ...
data/raw/batch-r02c.jsonl	2	{"id": "ttf-r02c-012", "title": "Samaco-Fell SF-3 / Press P-8: off-gas O2 84 ppm beats H2 2.4 Nm3/h by 168 us; correct MODIFY clamps hydrogen 2.4 -> 1.1", "state": {"description": "Samaco-Fell SF-3 binder-burn on Press P-8 publishes zirconia off-gas at 84 ppm O2, which is 44 ppm above the 40 ppm SmCo oxidation limit, w...
data/raw/batch-r02c.jsonl	3	{"id": "ttf-r02c-013", "title": "Tetrach-Germ TG-2 / Kettle K-9: AE 52 pps beats kettle 86 C by 176 us; correct REJECT holds GeCl4 1.6 -> 0 kg/h", "state": {"description": "Tetrach-Germ TG-2 hearts-cut on Kettle K-9 carries a live quartz-split signature: acoustic emission 52 pps versus a 10 pps crack trip, while the st...
data/raw/batch-r02c.jsonl	4	{"id": "ttf-r02c-014", "title": "Tantala-Lith TL-2 sim / Puller X-4: seed 1488 C beats furnace glint 1610 C by 178 us; correct ACCEPT leaves 0.22 mm/h", "state": {"description": "Tantala-Lith TL-2 is a lighting-glint simulation of Puller X-4: seed RTD 1488 C is 52 C under the 1540 C freeze-lid limit at a filed 0.22 mm/...
data/raw/batch-r02c.jsonl	5	{"id": "ttf-r02c-015", "title": "Oxychlor-Van OV-6 / Reactor R-11: live Cl2 0.42 vol% beats lagged XS-12 by 178 us; WRONG REJECT holds VOCl3 18 -> 0 (lagged interlock as trip)", "state": {"description": "Oxychlor-Van OV-6 finished CIP 840 s ago, yet dry-contact XS-12 on Reactor R-11 remains closed while live TCD chlori...
data/raw/batch-r03.jsonl	1	{"id":"ttf-r03-011","title":"Calyx-Brae CB-9 / Scope-U4: tip FT beats irrigation flowmeter by 320 us; correct MODIFY still eats an in-window stone roll (partnered negative total -0.42)","state":{"description":"Calyx-Brae phantom CB-9 already has Scope-U4's tip 18 mm past the ureteropelvic junction with live force 3.10 ...
data/raw/batch-r03.jsonl	2	{"id":"ttf-r03-012","title":"Roundel-Wath RW-5 / Ego-E7: radar gore occupancy beats camera TTC by 340 us; correct MODIFY clamps 14 -> 6 m/s","state":{"description":"Ego-E7 is already committed 6.4 m into the Roundel-Wath RW-5 yield lane when radar occupancy of the inscribed gore sits at 0.31 m2 against a 0.15 m2 cap. C...
data/raw/batch-r03.jsonl	3	{"id":"ttf-r03-013","title":"Tread-Kame TK-HIL / Biped-K4: stance CoP beats IMU tilt by 560 us; correct REJECT freezes the 0.42 m swing","state":{"description":"TK-HIL belt encoder at Tread-Kame already shows Biped-K4's stance CoP 0.14 m outside the 0.08 m polygon while IMU tilt is only 2.1 deg under a 6.0 deg trip. A ...

End of preview. Expand in Data Studio

Thalamic Relay Trajectories (Grok 4.6)

Rights & intended use: public research corpus, not training data. Hosted frontier-model outputs are research-only inputs under project policy (synthetic-factory#161): intended_use: research_only, project_training_policy: blocked. Not training data for any model-weight update. Machine-readable record: rights.json. License: Synthetic Factory Research-Only License v1.0 (license: other, see LICENSE) (non-commercial).

Release status: the raw, uncurated payload of the 2026-09-02 SuperGrok Heavy run is published under data/raw/. It is available for inspection, evaluation, and reproducibility, but it is not training-ready (release-status.json: release_stage: raw_uncurated_public, training_ready: false, strict_audit: pending_curation_integration).

Visibility: public raw-data repository.

Rights and intended use

This is a public research corpus, not training data.

Under the Synthetic Factory project-policy decision recorded in rmems/synthetic-factory#161, outputs of hosted frontier models (here: Grok 4.6 (xAI), generated in the SuperGrok Heavy consumer chat surface) are research-only inputs:

intended_use: research_only
project_training_policy: blocked

These records must not enter SFT, DPO, RL, distillation, continued pretraining, or any other model-weight update in this project. training_ready: false is a data-quality statement; even a future curated release would not make this corpus eligible for weight updates under project policy. That policy value does not change merely because a provider document later permits training — changing it would require a separate recorded project decision.

Research retention, evaluation, and redistribution rights are tracked separately per provider/channel/date. The owner's decision on #161 (2026-09-04) is that research-only outputs may be retained, evaluated, and redistributed for research while training stays blocked; the machine-readable statuses in rights.json nevertheless fail closed as unresolved until the provider terms snapshot is pinned (synthetic-factory#163). provider_training_status and weight_publication_status are unresolved, which resolves to blocked for any dependent activity.

This public raw release is licensed under Synthetic Factory Research-Only License v1.0 (license: other, see LICENSE). That research-only license grants reuse permissions; it does not make the records training-ready, and it does not lift the project's training block.

What this dataset is

Purpose-specific synthetic trajectories for relay-gated state assessment, proposal evaluation, safety disposition, reward attribution, and state update. Kind comes from the payload, not the slug: every published JSONL row is a thalamic-gate wrap record — a state, a proposed_action, a safety_decision (ACCEPT / MODIFY / REJECT), the executed_action, its future_outcome, and a reward_components decomposition — carrying the v2 additions (id, title, spike_events, raster, gate_snn) that the Fable-era rounds did not have.

The release contains 135 thalamic-gate wrap records across 27 JSONL files (~1.26 MB of raw JSONL), all stamped meta.factory: thalamic-trajectory-factory, meta.generator: grok-4.6, and meta.run_label: 2026-09-02-final-heavy. Record ids run ttf-r01-001 … ttf-r72-360 and are unique across the release. Gate outcomes: ACCEPT 34 / MODIFY 61 / REJECT 40. Every row shares the top-level keys executed_action, future_outcome, gate_snn, id, meta, proposed_action, raster, reward_components, safety_decision, spike_events, state, title.

This is the Grok 4.6 counterpart of rmems/thalamic-relay-trajectories (Claude Fable 5). The two datasets come from the same factory prompts and round protocol but from different generators and different runs; they live in separate repositories and separate collections and are never merged.

Provenance

Generator: xAI Grok 4.6, run in the SuperGrok Heavy consumer chat surface (meta.rights.subscription_plan: SuperGrok Heavy), final run 2026-09-02 (meta.run_label: 2026-09-02-final-heavy). Every record carries meta.generator: grok-4.6 and a per-record meta.rights block recording the provider, model, channel, plan, surface, generation timestamp, and record-level policy fields (135 of 135 rows also carry the individual research/redistribution status fields). Record-level generation_surface values in this release: SuperGrok Heavy chat: 135.
Method: Synthetic Data Factory prompts run in SuperGrok Heavy chat and published through the factory's round protocol; operator and review: Raul Montoya Cardenas; release engineering for this upload: Claude Fable 5.1 (Claude Code). Details in provenance.json.
Rounds published: r01, r02, r02c, r03, r04, r05, r21, r22, r22c, r23, r23c, r24, r41, r42, r43, r61, r62, r63, r63c, r64, r65, r66, r67, r68, r69, r70, r72.
Selection rule: batch-rNNc.jsonl files are additional batches of the same round; both are published. The only duplicate, batch-r21.jsonl of failure-as-fuel-preference-cascade (identical ids to batch-r21c.jsonl), is omitted. Every NOTES-r*.md written by the factory is published verbatim under notes/.
Source: the factory's immutable raw tree outputs/raw/2026-09-02-final-heavy/thalamic-trajectory-factory/. Round markers, reservation and completion markers, NEXT_ROUND.json, hidden files, and staging directories are not part of the release.
Gates: pipelines/validate_run.py and pipelines/check_records.py from rmems/synthetic-factory report 0 errors on exactly the published files; rights.json is accepted by pipelines/rights_document.py (schema 0.1.0).

Raw file	Records	SHA-256 (prefix)
`data/raw/batch-r01.jsonl`	5	`ab1d51f7c94b…`
`data/raw/batch-r02.jsonl`	5	`8e8a94fdc852…`
`data/raw/batch-r02c.jsonl`	5	`b0d895966a80…`
`data/raw/batch-r03.jsonl`	5	`721959b52e1d…`
`data/raw/batch-r04.jsonl`	5	`f06cff1c00f9…`
`data/raw/batch-r05.jsonl`	5	`677d5c6ec123…`
`data/raw/batch-r21.jsonl`	5	`e78b4086f2af…`
`data/raw/batch-r22.jsonl`	5	`849b0a7350c4…`
`data/raw/batch-r22c.jsonl`	5	`e72da9d2d838…`
`data/raw/batch-r23.jsonl`	5	`439d21f67cd1…`
`data/raw/batch-r23c.jsonl`	5	`8f5a1e126ed9…`
`data/raw/batch-r24.jsonl`	5	`5029760f19c6…`
`data/raw/batch-r41.jsonl`	5	`9f8d7ac52ccd…`
`data/raw/batch-r42.jsonl`	5	`2fba07345881…`
`data/raw/batch-r43.jsonl`	5	`3af0436aa515…`
`data/raw/batch-r61.jsonl`	5	`741f155147a7…`
`data/raw/batch-r62.jsonl`	5	`925b0c2a3be0…`
`data/raw/batch-r63.jsonl`	5	`2e7be69b70d8…`
`data/raw/batch-r63c.jsonl`	5	`56915dbe5a53…`
`data/raw/batch-r64.jsonl`	5	`cb19b2af5635…`
`data/raw/batch-r65.jsonl`	5	`28854f9446c5…`
`data/raw/batch-r66.jsonl`	5	`bda0a3851ab1…`
`data/raw/batch-r67.jsonl`	5	`2288d2691ed4…`
`data/raw/batch-r68.jsonl`	5	`b8ab94b24cb7…`
`data/raw/batch-r69.jsonl`	5	`017a65aed67b…`
`data/raw/batch-r70.jsonl`	5	`51a32cbd7955…`
`data/raw/batch-r72.jsonl`	5	`86a86c311764…`

Full digests are in provenance.json raw_snapshot.

How to read

data/raw/batch-r*.jsonl is the source of truth: each file is a byte-for-byte copy of the selected factory file, one JSON record per line. Do not rewrite it; do not point a default config at data/raw/*.jsonl (nested key-bags drift across rounds).
data/viewer/records.parquet is a lossless viewer projection with three columns: source_file (the repo-relative raw path), source_line (1-based physical line) and record_json (the verbatim JSON line). Joining a file's record_json values in source_line order with \n reproduces the raw file byte for byte. The Dataset Viewer config viewer / split train indexes 135 rows.
notes/NOTES-r*.md are the factory's per-round notes, verbatim.
rights.json, release-status.json, and provenance.json are the machine-readable rights, release, and provenance sidecars.

import json
import pyarrow.parquet as pq

table = pq.read_table("data/viewer/records.parquet")
record = json.loads(table.column("record_json")[0].as_py())
print(record["id"], record["meta"]["generator"], record["meta"]["round"])

Do not train on any field of this raw Hub copy; it is evidence for research, evaluation, and cross-generator comparison only.

Relationship to the Fable dataset

rmems/thalamic-relay-trajectories holds the Claude Fable 5 rounds of the same factory; this repository holds the Grok 4.6 rounds of the 2026-09-02 SuperGrok Heavy run. Same family, same schema lineage, different generator, different run, different license (the Fable release is Apache-2.0; this one is Synthetic Factory Research-Only License v1.0 (license: other, see LICENSE)). Grok output is never pushed into the Fable repository and the two are never merged; comparing them is one of the intended research uses.

License

This public raw release is licensed under the Synthetic Factory Research-Only License v1.0 (license: other, license_name: synthetic-factory-research-only) (Synthetic Factory Research-Only License v1.0 (license: other, see LICENSE)). That license grants non-commercial reuse permissions; it does not make the records training-ready or factual real-world measurements, and project policy blocks training use regardless of license.