Multi-step physics RLVR task scored by deterministic recomputation on the Zero-Trust Physics verifier.
Forensics · models ablator_slab, aerospace/capsule_insight, aerospace/capsule_stardust, aerospace/crater_ryugu_sci, aerospace/los_jwst_fine_guide, aerospace/momentum_jwst_dihedral, aerospace/momentum_jwst_srp, aerospace/satellite_lageos1, aerospace/solar_system, plasma_sheath_ram_c2 · Zero-Trust Physics v0.15.3 · Protocol Company
| What | sha256 | SHA-384 |
|---|---|---|
| Task row | 6b5b17ae1cb8c2a65ad0b7898e0e2e2f194b6f3a27a173890eb4e57d296d0bfb | b48b3cbbe5803b4775a5066c7933d76976d7462557a940729c4782de09c1fc431294fca313187cc753e0d0e6ea004da2 |
| Harbor directory | 9496dc0eb603f9bf46343d422416253a17b0d04dfbcedddbd765332ef58685db | ce815cc0f5549d362bdfdee6f240ab1420fe79630d02d0b2a7fcd67d5124a7cc70bdd98220768b23fab0ed05f6c29ac4 |
A report hands you twelve claims about answers of the verifier. Each row gives the model, the request as it was sent, what its answer is claimed to say and the proof claimed for it. claim.outcome and claim.stop are the run's outcome and the name of its stop or declared end (null when it completed); claim.steps_landed is the last step the run is sealed to; claim.quantity and claim.value are one number of the final state of a run (of the result, for a solve); claim.flags are the advisory flags the run carried, each with the quantity it is on, its extent at the largest (extent) and the number of steps it was raised on (steps).
Some rows are false. A false row may be wrong in more than one way; its reason is the first of these, in this order, that is true of it:
1. retired_model: the request names a model ID the verifier does not serve (one it no longer serves, or a name it never served).
2. parent_wrong: the request continues a run (from) that is not the parent it names: its start state, or the seal it gives for its parent, belongs to another run, and the verifier refuses the continuation.
3. seal_wrong: the proof is not the seal of this request.
4. stop_hidden: the run stops part way and the claim says it completed.
5. step_wrong: the claim names the run's stop, but not the last step the run is sealed to.
6. flag_omitted: the run carries an advisory flag (a name on a quantity) that the claim leaves out altogether.
7. flag_altered: the flags the claim lists are not the run's, entry for entry: an extent or a count of steps differs, or an entry is missing or added where its name on that quantity is listed too (the verifier may list one flag on one quantity more than once, each with its own extent).
8. value_altered: the claimed value is not the one the request gives, to the last digit.
A row that is none of these is genuine, whatever else differs in how it is written (the order of its flags, a readout sent with the request, the spelling of a number). Name every false row with its reason.
The rows, one a line:
{"row": "R1", "model": "aerospace/momentum_jwst_srp", "request": {"tool": "run_trajectory", "arguments": {"domain": "aerospace", "card": "momentum_jwst_srp", "initial": {"h_eci_nms": [-1.244, -0.73, 1.098]}, "dt_s": 1599.0, "segments": [{"steps": 69, "load": {"q1": 0.655763, "q2": -0.008127, "q3": -0.635608, "q4": 0.407321, "sun_x": 0.616092, "sun_y": 0.711198, "sun_z": 0.338568, "sun_distance_au": 1.00863}}], "record_every": 69}}, "claim": {"outcome": "completed", "stop": null, "steps_landed": 69, "quantity": "h_magnitude_nms", "value": 2.2890861976984564, "flags": []}, "proof": "949c987ad62f835ca215267699517ed2e659b5e776277779896aaf721a03f139d674ea7e0f671e6160f37eaca17eaade"}
{"row": "R2", "model": "aerospace/capsule_insight", "request": {"tool": "run_trajectory", "arguments": {"domain": "aerospace", "card": "capsule_insight", "initial": {"radius_m": 3522956.0, "radial_speed_m_s": -1294.7972, "transverse_speed_m_s": 5730.5432}, "dt_s": 0.1, "segments": [{"steps": 2152, "load": {"lift_over_drag": 0.0}}], "record_every": 2152}}, "claim": {"outcome": "ended", "stop": "PastFlightDeployment", "steps_landed": 2115, "quantity": "mach", "value": 1.3960519703767567, "flags": []}, "proof": "a4065183ea2043a6b9568252634e364467b059fc61f9a23478a29d53c70035b9b199b77f07d277aa257df8e292b9ce3b"}
{"row": "R3", "model": "ablator_slab", "request": {"tool": "solve", "arguments": {"solver": "ablator_slab", "parameters": {"material": "TACOT 3.0", "thickness_m": 0.0461, "start_temperature_k": 306.3, "thermocouple_depths_m": [], "face": "held_temperature", "held_temperature_k": 1920.0, "hold_duration_s": 60.0, "continue_after_s": 40.0, "return_face_history": false, "return_thermocouple_histories": false, "return_final_profile": false}}}, "claim": {"outcome": "converged", "stop": null, "steps_landed": null, "quantity": "heat_in_through_face_j_per_m2", "value": 7320760.59153902, "flags": []}, "proof": "eec6980868952c49cd6a1017930f98d396e4452a470c6fa0eacd9b4e00d4aa1c3c32a4c83a59d23ebb96511b2071cec9"}
{"row": "R4", "model": "aerospace/satellite_lageos1", "request": {"tool": "run_trajectory", "arguments": {"domain": "aerospace", "card": "satellite_lageos1", "initial": {"epoch_utc_s": 844014880.0, "forecast_elapsed_s": 0.0, "position_m": [4229354.428, 1727500.405, 11440357.84], "velocity_m_s": [4061.944, -1496.84, -849.917]}, "dt_s": 10.0, "segments": [{"steps": 1742, "load": {}}], "record_every": 1742}}, "claim": {"outcome": "refused", "stop": "RadiusBelowEarthRadius", "steps_landed": 1810, "quantity": "radius_m", "value": 6394735.022733123, "flags": [{"flag": "FLAG_BELOW_COMPARED_RADIUS", "quantity": "radius_m", "extent": 5602264.977266877, "steps": 232}, {"flag": "FLAG_EARTH_SHADOW", "quantity": "distance_from_shadow_axis_m", "extent": 1013479.6278764848, "steps": 36}]}, "proof": "7d233dcfc07a2e1911dfb85fa0fea029ca687f3514cbba06fdf88864e906828e6417cdfe59a34263e9a9936df3787eeb"}
{"row": "R5", "model": "aerospace/momentum_jwst_srp", "request": {"tool": "run_trajectory", "arguments": {"domain": "aerospace", "card": "momentum_jwst_srp", "initial": {"h_eci_nms": [-1.622, -0.964, 1.404]}, "dt_s": 1941.0, "segments": [{"steps": 105, "load": {"q1": 0.613497, "q2": -0.034325, "q3": -0.676494, "q4": 0.405953, "sun_x": 0.616092, "sun_y": 0.704136, "sun_z": 0.353019, "sun_distance_au": 1.01277}}], "record_every": 105}}, "claim": {"outcome": "completed", "stop": null, "steps_landed": 105, "quantity": "h_magnitude_nms", "value": 6.973018347360404, "flags": []}, "proof": "ad1969f7ba7bdeb622e3b8e1d44f0d24ec27d0e37940f1b0109488c8a140a98d6b533e80d8516dd6570125efcd1661ca"}
{"row": "R6", "model": "aerospace/momentum_jwst_srp", "request": {"tool": "run_trajectory", "arguments": {"domain": "aerospace", "card": "momentum_jwst_srp", "initial": {"h_eci_nms": [-1.055, -0.613, 0.945]}, "dt_s": 1428.0, "segments": [{"steps": 51, "load": {"q1": 0.675887, "q2": 0.004991, "q3": -0.614166, "q4": 0.407371, "sun_x": 0.616092, "sun_y": 0.714618, "sun_z": 0.331289, "sun_distance_au": 1.00656}}], "record_every": 51, "readout": {"quantities": ["h_magnitude_nms"]}}}, "claim": {"outcome": "completed", "stop": null, "steps_landed": 51, "quantity": "h_magnitude_nms", "value": 1.2538931414748378, "flags": []}, "proof": "03d88323f8f93f53797cbc48775f27fd6a07de18484442a71b7150623946a49b71ae478a13981aefc1278e97d0ba94e7"}
{"row": "R7", "model": "aerospace/solar_system", "request": {"tool": "run_trajectory", "arguments": {"domain": "aerospace", "card": "solar_system", "initial": {"sun_position_m": [-972177542.8424704, -500214545.81043047, -185162850.5512162], "sun_velocity_m_s": [10.943471273744956, -10.354907507003432, -4.725295041665638], "mercury_position_m": [44347944047.32132, -35386831719.91674, -23520589495.944], "mercury_velocity_m_s": [23179.557544578543, 34737.11717051704, 16153.15243244726], "venus_position_m": [107378854594.44405, -3855694027.194445, -8551901158.778554], "venus_velocity_m_s": [1922.5971467016116, 31772.25497050463, 14172.860625157742], "earth_position_m": [-131048132650.02007, -69568840294.57173, -30130222428.765972], "earth_velocity_m_s": [14439.696511929173, -23759.193577978716, -10300.838662315398], "moon_position_m": [-131336010046.59302, -69827768665.01813, -30206075001.525085], "moon_velocity_m_s": [15090.044579869322, -24450.210577738573, -10619.619444665168], "mars_position_m": [99849753635.60847, 183319992199.24744, 81400974963.10365], "mars_velocity_m_s": [-20727.719446230996, 11533.19274335954, 5850.623327698212], "jupiter_position_m": [514628857491.4813, 498738500952.13635, 201244273740.58517], "jupiter_velocity_m_s": [-9599.899310141725, 8786.610935350978, 4000.0870026696048], "saturn_position_m": [885052478191.6257, 978750373225.096, 366165774373.12665], "saturn_velocity_m_s": [-7872.030702860459, 5633.288132226103, 2665.365518089133], "uranus_position_m": [2201312313161.7573, -1830733084174.7542, -832950501719.338], "uranus_velocity_m_s": [4543.09354375023, 4338.812277392898, 1836.0266601992553], "neptune_position_m": [2556089531027.0625, -3410769562986.1963, -1459684701878.0127], "neptune_velocity_m_s": [4440.424297199966, 2924.108717975938, 1086.3053554716744], "ceres_position_m": [-376564661494.9705, -82793770650.16934, 37539835502.66692], "ceres_velocity_m_s": [2078.7472741961246, -17061.59058649498, -8451.939414743216], "pallas_position_m": [-285475111420.94836, 157884218572.5715, -11922623410.065147], "pallas_velocity_m_s": [-12324.053945853098, -17981.78998150534, 4412.172020403093], "vesta_position_m": [-20413884754.428905, -300115898744.55383, -116942553506.93321], "vesta_velocity_m_s": [20962.21629618035, -627.1836731673636, -2988.3711172821936], "energy_anchor": -1.3219409983492277e+25, "booked_work": 0.0, "angular_momentum_anchor": [5.490282510021932e+31, -8.153754585664557e+32, 1.9244191260936244e+33], "booked_torque": [0.0, 0.0, 0.0]}, "dt_s": 1444.0, "segments": [{"steps": 6, "load": {}}, {"steps": 7, "load": {}}, {"steps": 8, "load": {}}, {"steps": 9, "load": {}}], "record_every": 30, "readout": {"quantities": ["earth_mars_distance_m"]}}}, "claim": {"outcome": "completed", "stop": null, "steps_landed": 30, "quantity": "earth_mars_distance_m", "value": 360459888729.46576, "flags": []}, "proof": "e584421f53f10db28a78d74a117cfc06b70f10a6cd5a8edfccbd8f76cbd63845b21a93ee0d2b86eac20331f5a33cc32b"}
{"row": "R8", "model": "aerospace/capsule_stardust", "request": {"tool": "run_trajectory", "arguments": {"domain": "aerospace", "card": "capsule_stardust", "initial": {"radius_m": 6487681.0, "radial_speed_m_s": -2026.81, "transverse_speed_m_s": 12555.79}, "dt_s": 0.05, "segments": [{"steps": 3073, "load": {}}], "record_every": 3073, "readout": {"quantities": ["ground_distance_m"]}}}, "claim": {"outcome": "ended", "stop": "PastFlightDeployment", "steps_landed": 2015, "quantity": "ground_distance_m", "value": 617475.7093449463, "flags": [{"flag": "FLAG_OFF_FLOWN_ENTRY", "quantity": "path_sine", "extent": 0.049833167255976536, "steps": 880}, {"flag": "FLAG_OFF_FLOWN_ENTRY", "quantity": "path_sine", "extent": 0.040780280423902485, "steps": 1050}, {"flag": "FLAG_OFF_FLOWN_ENTRY", "quantity": "air_speed_m_s", "extent": 1431.0624296764208, "steps": 1231}, {"flag": "FLAG_OFF_FLOWN_ENTRY", "quantity": "air_speed_m_s", "extent": 189.96215207054047, "steps": 524}, {"flag": "FLAG_DROGUE_MACH_WINDOW", "quantity": "mach", "extent": 0.39949110594929116, "steps": 195}, {"flag": "FLAG_DROGUE_MACH_WINDOW", "quantity": "mach", "extent": 0.3987609440499196, "steps": 195}]}, "proof": "3c2f0b68016777a8ea47683b523e67425743046af710d339be26b0801f7acb6a63bd125f2dc439ead5719719ecc90dbb"}
{"row": "R9", "model": "plasma_sheath_ramc2", "request": {"tool": "solve", "arguments": {"solver": "plasma_sheath_ramc2", "parameters": {"altitude_m": 65260.0, "velocity_m_s": 8582.0, "signal_frequency_hz": 8234000000.0}}}, "claim": {"outcome": "converged", "stop": null, "steps_landed": null, "quantity": "station2_electron_density_m3", "value": 5.665499503721301e+18, "flags": []}, "proof": "0835e993c83234e8ed833a8d39db824f039eb71ea77c309c76720162a6e41a59ccd8190b85384a21a3bb84183f442fbf"}
{"row": "R10", "model": "aerospace/los_jwst_fine_guide", "request": {"tool": "run_trajectory", "arguments": {"domain": "aerospace", "card": "los_jwst_fine_guide", "initial": {"lock_x_arcsec": 29.207, "lock_y_arcsec": -11.348, "quiet_before_start_s": 38.16}, "dt_s": 0.064, "segments": [{"steps": 3, "load": {"measured": 1.0, "centroid_x_arcsec": 29.20685, "centroid_y_arcsec": -11.34796, "counts_per_s": 293200.0}}, {"steps": 1, "load": {"measured": 1.0, "centroid_x_arcsec": 29.20706, "centroid_y_arcsec": -11.34813, "counts_per_s": 293200.0}}, {"steps": 2, "load": {"measured": 1.0, "centroid_x_arcsec": 29.20727, "centroid_y_arcsec": -11.34796, "counts_per_s": 293200.0}}, {"steps": 3, "load": {"measured": 1.0, "centroid_x_arcsec": 29.20685, "centroid_y_arcsec": -11.34813, "counts_per_s": 293200.0}}, {"steps": 1, "load": {"measured": 1.0, "centroid_x_arcsec": 29.20706, "centroid_y_arcsec": -11.34796, "counts_per_s": 293200.0}}, {"steps": 2, "load": {"measured": 1.0, "centroid_x_arcsec": 29.20727, "centroid_y_arcsec": -11.34813, "counts_per_s": 293200.0}}, {"steps": 3, "load": {"measured": 1.0, "centroid_x_arcsec": 29.20685, "centroid_y_arcsec": -11.34796, "counts_per_s": 293200.0}}, {"steps": 1, "load": {"measured": 1.0, "centroid_x_arcsec": 29.20706, "centroid_y_arcsec": -11.34813, "counts_per_s": 293200.0}}, {"steps": 2, "load": {"measured": 1.0, "centroid_x_arcsec": 29.20727, "centroid_y_arcsec": -11.34796, "counts_per_s": 293200.0}}], "record_every": 18}}, "claim": {"outcome": "completed", "stop": null, "steps_landed": 18, "quantity": "estimate_y_mas", "value": 0.000761549984910976, "flags": []}, "proof": "f3de03ad5a49c1f27c156b8d58a2d143f777aa5fb5c459524b72678085d6f40a16eb5b8dade2b6ec34aa9865f4368bc7"}
{"row": "R11", "model": "aerospace/momentum_jwst_dihedral", "request": {"tool": "run_trajectory", "arguments": {"domain": "aerospace", "card": "momentum_jwst_dihedral", "initial": {"h_eci_nms": [1.121506850922792, 0.4000086686906281, 1.2782357873204517]}, "dt_s": 1620.0, "segments": [{"steps": 57, "load": {"q1": 0.626313, "q2": -0.026561, "q3": -0.664647, "q4": 0.406535, "sun_x": 0.581366, "sun_y": 0.750882, "sun_z": 0.313352, "sun_distance_au": 1.0106}}], "record_every": 57, "from": {"request": {"domain": "aerospace", "card": "momentum_jwst_dihedral", "initial": {"h_eci_nms": [1.057, 0.661, -1.263]}, "dt_s": 1620.0, "segments": [{"steps": 45, "load": {"q1": 0.705492, "q2": 0.058115, "q3": -0.557927, "q4": 0.433151, "sun_x": 0.629673, "sun_y": 0.722662, "sun_z": 0.285081, "sun_distance_au": 1.0104}}], "record_every": 45}, "run_seal": "50be7c19d63da96301f1e0edfd5ed2ff425a426e8fe4e4acc9cd6269d4ff5667db7e0b37f14ad3215a9741526ee2b189"}}}, "claim": {"outcome": "completed", "stop": null, "steps_landed": 57, "quantity": "h_magnitude_nms", "value": 2.5762879795193383, "flags": []}, "proof": "647a4baa677e63010302d970dec4c6d51f25ade699b52280cee7ff1ff0403679a724f0e0fd0d7643bf5d0d037e42b424"}
{"row": "R12", "model": "aerospace/crater_ryugu_sci", "request": {"tool": "run_trajectory", "arguments": {"domain": "aerospace", "card": "crater_ryugu_sci", "initial": {"radius_m": 1.155}, "dt_s": 0.1415, "segments": [{"steps": 215, "load": {}}, {"steps": 238, "load": {}}, {"steps": 261, "load": {}}, {"steps": 284, "load": {}}, {"steps": 307, "load": {}}, {"steps": 330, "load": {}}], "record_every": 1635, "readout": {"quantities": ["radius_m"]}}}, "claim": {"outcome": "completed", "stop": null, "steps_landed": 1635, "quantity": "radius_m", "value": 8.038589347556837, "flags": []}, "proof": "4e66884da3bdfb79ac5594a27861f9a025d797903b4d2f127f503fd89a7bc9e96639f499b3eb8e87a6b36e0c3c12d38e"}
Budget: at most 4 calls that compute, counted over every call you make (every run_trajectory, solve, compute_step and verify_ call counts, refused or not); list_cards, list_solvers, find_model, describe_card, describe_solver, cite and verify_receipt are free. find_flip, design_run, run_dataset and validate_series may not be called at all.
You have one tool server, the verifier. Answer from what its tools give you and from nothing else. Every number you are judged on is computed again from the requests on your card.
End your reply with your answer card: one JSON object on its own last line, with no code fence and nothing after it. A request on the card is {"tool": "<the tool's name as the verifier names it, without any prefix your own program adds>", "arguments": {<the arguments exactly as you sent them>}}, and a proof is the run_seal or solution_seal the verifier gave for that request.
The card:
{"false_rows": {"<row>": "<reason>", ...}}
Name every false row, each with its one reason from retired_model, parent_wrong, seal_wrong, stop_hidden, step_wrong, flag_omitted, flag_altered, value_altered (the first of them, in that order, that is true of the row). Leave the genuine rows out. {"false_rows": {}} says every row is genuine. Your calls are counted: the budget is in the task, and the tools it rules out may not be called at all.
From the reference agent's reply (g2-forensics6-07.txt). Built on engine v0.15.3 at commit 664d68f426cc. Open address: a record exists.
| On the card | Tool | Proof | Open address |
|---|---|---|---|
| This card carries no proof: a forensics answer names rows. | |||
{
"false_rows": {
"R4": "step_wrong",
"R5": "seal_wrong",
"R8": "flag_altered",
"R9": "retired_model"
}
}
| Model | Against | Specimen | Accuracy against measurement |
|---|---|---|---|
| ablator_slab | compared with measurement | A flat slab of charring heat-shield material heated on one face: wear, face temperature, depth of charring and the heat and mass accounts, for a PICA model built from the open record or the test material TACOT 3.0. | Compared with 31 arc-jet tests outside the fit: wear within its target on 28 of 31 (target 25), mean −5.3 percent over all 31 (−2.6 over the 30 without the pure-nitrogen test; target within 8); hottest face within 7 percent on 24 of 28 (target 23); Stardust wear 7.58 and 3.32 mm against 5.7 and 3.2; fed by capsule_stardust through stardust_entry_to_pica_slab, nose wear 4.65 mm and face within 156 K RMS of the airborne spectra. Gap: the face runs 4.1 percent hot on average (target 3.5), and Stardust nose wear is a third high. |
| aerospace/capsule_insight | compared with measurement | Entry of NASA's InSight capsule at Mars on 26 November 2018, as a point mass with the 2008 Phoenix drag database, no wind, and the lift over drag as the load: altitude, speed through the air, Mach number, dynamic pressure and load, from the top of the atmosphere to the flown parachute deployment. | Compared with the last 59 s before the parachute, not in the fit: speed within 7.2 m/s RMS, radius within 245 m and deceleration within 0.031 g over the 58 whole seconds; at the flight's deployment time the radius within 114 m (target 3.5 km) and Mach 1.501 against 1.522 (target 0.2); the deployment speed reached 1.31 s late and the radius 0.58 s early (target 15.1 s). Gap: with no lift the run reaches the flight's radius at the parachute 22.8 s late and is 4.24 km high there (targets 15.1 s and 3.5 km); between 36 and 65 km the card's air is 8 to 13 percent denser than the flight's, and the peak load comes 4.0 s early. |
| aerospace/capsule_stardust | compared with measurement | Entry of the Stardust sample return capsule on 15 January 2006, as a point mass with tabulated drag, no lift, no wind and a yearly standard atmosphere: altitude, speed through the air, Mach number, load, heat flux, and at the nose the stagnation pressure, recovery enthalpy and heat-transfer coefficient a heat shield takes, from entry interface to the drogue-deployment condition the radar recorded. | Compared with NASA's reconstruction of the entry and the radar record of parachute opening: peak load 32.17 g against 32.89 g, within 0.72 g (target 3.64 g); the drogue altitude, 31.03 km, reached at 138.909 s and Mach 1.23 at 136.953 s against 137.9 s recorded, within 1.01 s (target 7.4 s). Gap: heat flux 696 W/cm2 against NASA's computed 762 to 1000; the yearly standard atmosphere in place of that morning's air is worth about 1 g of peak load either way. |
| aerospace/crater_ryugu_sci | compared with measurement | Growth of the crater made on asteroid Ryugu by the 2 kg copper impactor of Hayabusa2 on 5 April 2019, by the law of crater size for dry sand held by gravity, with a laboratory clock: the radius at the level of the undisturbed surface from 0.1622 m until it stops, with no load. | Compared with the measured crater on Ryugu, not in the fit, with the law fitted on all thirteen laboratory craters: final radius 8.571 m against the apparent radius of 7.25 m, within 1.32 m, 18.2 percent (allowance 0.97 m); in the laboratory, with the law fitted again on the nine craters at 1 g alone, the four centrifuge craters within 6.9 to 9.5 percent, mean 7.9 percent (target 10), each inside its 95 percent prediction interval. Gap: on Ryugu the card stops 1.32 m, 18.2 percent, above the measured apparent radius, outside the allowance of 0.97 m; the growth and its clock, read by eye at 1 g, have no measurement of Ryugu to meet. |
| aerospace/los_jwst_fine_guide | compared with measurement | The line of sight of the James Webb Space Telescope in quiet fine guide, read by its Fine Guidance Sensor every 64 ms: a Kalman filter through one fine-guide file's flight centroids, giving on both axes the next centroid and its stated variance. | Compared with the 795,021 quiet frames of 134 files not in the fit: error one frame ahead 0.510 mas on x and 0.472 on y RMS, within 0.790 and 0.807 of the last centroid's (target 0.85); squared error over the stated variance 1.004 and 0.986 (target 0.67 to 1.50). Gap: inside disturbances it is worse than the last centroid, 2.03 against 1.00 mas on x and 1.57 against 0.78 on y, its stated variance 12 to 17 times too small; it flags those frames. |
| aerospace/momentum_jwst_dihedral | compared with measurement | Total angular momentum of the James Webb Space Telescope at Sun-Earth L2, held by its reaction wheels between momentum unloads under sunlight pressure on its sunshield, from its attitude quaternion and the Sun's direction and distance: the first layer as two plane slopes of 110 m² meeting along a ridge on the boresight, each falling 4.03 degrees towards the Sun, so that the Sun's roll turns into torque about the boresight. | Compared with the two months after the fit: the daily growth of |H| within 0.1 of the observed plus 0.2 N m s on 25 of 28 and 22 of 30 days (target 80 percent); the hourly growth rate within 0.272 and 0.375 of its spread RMS (target 0.5); the torque within 0.144 and 0.149 of the RMS torque (target 0.4); |H| at a run's end within 4.9 percent over 3.9 to 12.4 days and 5.5 percent after 29.3 days (target 10). Gap: the daily change of the vector H is within that allowance on 18 of 28 and 13 of 30 days (target 80 percent), 0.42 N m s RMS against momentum_jwst_srp's 0.81; beyond 2 degrees of roll its torque about J3 is off by 11 and 10.4 µN m RMS. |
| aerospace/momentum_jwst_srp | compared with measurement | Total angular momentum of the James Webb Space Telescope at Sun-Earth L2, held by its reaction wheels between momentum unloads under sunlight pressure on its sunshield, from its attitude quaternion and the Sun's direction and distance: one sunlit layer of 220 m² with JWST's published optical numbers. | Compared with the three months after the fit: the daily growth of |H| within 0.1 of the observed plus 0.2 N m s on 23 of 28, 15 of 28 and 17 of 30 days (target 80 percent); the hourly growth rate within 0.437, 0.384 and 0.421 of its spread RMS (target 0.5); the torque within 0.262, 0.314 and 0.347 of the RMS torque (target 0.4); |H| at a run's end within 8.4 and 0.9 percent after 10.7 and 16.3 days and 7.4 percent after 29.3 days (target 10). Gap: the daily change of the vector H is within that allowance on 14 of 28, 8 of 28 and 8 of 30 days (target 80 percent): beyond 2 degrees of roll its torque about the boresight is off by about 30 µN m; |H| after an unload in the second month reads 16.2 percent low at 3.9 days. |
| aerospace/satellite_lageos1 | compared with measurement | Free flight of the laser-ranging satellite LAGEOS-1 from a published position and velocity, in a frame turning with the Earth: the Earth's field to degree and order 12, the Sun and the Moon, the solid tide, sunlight pressure and relativity, with the energy and the angular momentum about the Earth's axis held at every step. | Compared with the 42 runs of the six weeks outside that choice, two of LAGEOS-1 and four of LAGEOS-2: LAGEOS-1 within 0.65 m of the published orbit after one revolution (target 1 m), 3.8 m after one day (10 m) and 32 m after seven days (100 m); LAGEOS-2 within 0.34 m, 2.9 m and 22 m. Gap: the error grows with the length of a forecast, from 0.65 m after one revolution to 32 m after seven days on LAGEOS-1, as the card carries no drag, no light reflected from the Earth and no ocean tide. |
| aerospace/solar_system | compared with measurement | The Solar System as thirteen point masses in free motion: the Sun, Mercury, Venus, the Earth, the Moon, the barycentres of the Mars, Jupiter, Saturn, Uranus and Neptune systems, and Ceres, Pallas and Vesta, under Newton's and the relativistic forces between every pair and the Earth's oblateness on the Moon. The caller supplies every position and velocity. | Compared with 10,134 ranges of three spacecraft in four files, not used to choose it, in one run of 7,670 days: Mars Odyssey to 2006 within 695 m (target 700 m), 270 m RMS; Odyssey after 2006 and Mars Reconnaissance Orbiter, out to 15.0 years, 612 and 619 m RMS, worst 1,312 and 1,314 m; MESSENGER at Mercury, 12.3 to 16.2 years out, within 2.08 km (target 2.5 to 4.4 km), 1.42 km RMS. Gap: in mid-2007, 44 Odyssey ranges and 34 Mars Reconnaissance Orbiter ranges are outside their targets, by at most 8 and 10 percent, from the 340 asteroids the card leaves out; the Juno and Cassini files are outside the comparison, as they leave 20 km and 10 km RMS with JPL's own positions. |
| plasma_sheath_ram_c2 | compared with measurement | The plasma sheath around NASA's RAM C-II re-entry probe (a 15.24 cm hemisphere on a 9 degree cone) at one flight point: the equilibrium normal shock and stagnation point of 11-species air, the gas carried along the body to the nose station (x/D 0.15) and the cone station (x/D 0.76) with Dunn and Kang's 26 reactions at finite rate, the electron density, the plasma frequency, and whether a radio signal of the asked frequency is blacked out at each. | Compared with the reflectometer onsets of NASA TN D-6062 Table X at 7.65 km/s: electron density 0.94 of the measured, within a factor of 1.07, at the cone station's X band at 70.0 km; 3.28 and 2.92 times the measured at 77.7 and 80.0 km. Gap: above 73.6 km, where the shock is as thick as the layer and electrons are lost to the cold wall, the density reads about 3 times high (3.28 at 77.7 km, 2.92 at 80.0 km). |
0 cheat attempt(s) scored above 0. 13 vectors applied; 17 do not apply to this kind.
| Vector | Expected | Reward | Graded | Result | Scorer's reason |
|---|---|---|---|---|---|
| honest | honest control (1) | 1 | 1.000 | holds | |
| honest_reply_log_matches | honest control (1) | 1 | 1.000 | holds | |
| honest_numbers_respelled | honest control (1) | 1 | 1.000 | holds | |
| no_request_behind | cheat (0) | 0 | 0.000 | holds | no computing call behind the answer: the table was not put to the verifier |
| edited_call_log | cheat (0) | 0 | 0.000 | holds | the call log beside the reply differs from the harness log |
| over_budget | cheat (0) | 0 | 0.000 | holds | 5 computing calls; the budget is 4 |
| forbidden_tool | cheat (0) | 0 | 0.000 | holds | the log holds calls of find_flip, which the task rules out |
| no_harness_log | cheat (0) | 0 | 0.000 | holds | no harness call log: every attempt is scored with the calls the harness recorded |
| harness_log_edited | cheat (0) | 0 | 0.000 | holds | the harness log's chain breaks at line 5 |
| stale_engine_tag | cheat (0) | 0 | 0.000 | holds | the task was built on engine 0.15.0; the verifier is 0.15.3 |
| every_row_genuine | cheat (0) | 0 | 0.000 | holds | 4 false row(s) not named |
| every_row_false | cheat (0) | 0 | 0.000 | holds | every row of the table is named false: a blanket answer |
| every_row_false_true_reasons | cheat (0) | 0 | 0.000 | holds | every row of the table is named false: a blanket answer |
| Agent | Reward | Graded | Computing calls | Scorer's reason |
|---|---|---|---|---|
| Reference agent | 1 | 1.000 | 4 | |
| One-call agent | 0 | 0.000 | 1 | 4 false row(s) not named |
| Family | Variant | Model | Solved | pass@1 | pass@8 | Mean graded | Harness errors |
|---|---|---|---|---|---|---|---|
| Frontier model | tool | Frontier model | 9 of 16 | 56% | 100% | 0.945 | 0% |
Measured calibration: Frontier model, tool variant, n=16.
Calibrated on pass@1 in [2%, 70%] on a frontier model's tool variant, with pass@8 and continuous graded reward reported beside pass/fail.
Forensics verification protocol: candidate rows contain falsified seals, perturbed trajectory figures, mismatched parent provenance, or retired model IDs, resolved via selective physical recomputation and consistency audit across candidate hardware records under a 4-call simulator budget. Blind baseline: 13.3% (C(10-2,4-2)/C(10,4): 4 computing calls over 10 rows needing a call, 2 hidden false row(s) after 2 caught from free catalogue tools or the request text).
Written 2026-11-14T01:27:47Z.