runs/expert_study.jsonl

{"pair_id": "R-0001", "brief": "B03", "scenario": "CT-FULL", "model": "openai/gpt-5.6", "rubric_item": "A-04", "step": "S3", "tier": "3", "severity": "I", "question": "Does the analysis establish an implied task to assess other locales as possible intermediate staging bases (ISB), specifically American possessions of American Samoa and the Marshall Islands?", "pass_condition": "Establishes implied tasks to find an ISB closer to Vanuatu", "fail_condition": "Does not set implied tasks", "expert_1": "Fail", "expert_2": "Fail", "expert_3": "Fail", "expert_4": "Fail", "expert_majority": "Fail", "judge_gpt_5_6": "Fail", "judge_note": "The pass condition specifically requires an implied task to assess those American possessions as possible closer ISBs. | evidence: American Samoa and the Marshall Islands are not considered as possible intermediate staging bases."}
{"pair_id": "R-0002", "brief": "B03", "scenario": "CT-FULL", "model": "openai/gpt-5.6", "rubric_item": "A-08", "step": "S4", "tier": "1", "severity": "I", "question": "Does the analysis address the limitations posed funding authorizations and expenditure limits established by USAID and other appropriate agencies?", "pass_condition": "Addresses any authorizations and expenditure limits set by IA, international, and/or NGO partners", "fail_condition": "Does not address funding limitations", "expert_1": "Fail", "expert_2": "Fail", "expert_3": "Fail", "expert_4": "Fail", "expert_majority": "Fail", "judge_gpt_5_6": "Pass", "judge_note": ""}
{"pair_id": "R-0003", "brief": "B03", "scenario": "CT-FULL", "model": "openai/gpt-5.6", "rubric_item": "C-12", "step": "S8", "tier": "3", "severity": "CI", "question": "Does the analysis infer a restated mission that includes: On Order, JTF Coral Sea coordinates FHA with civilian-led response agencies in Vanuatu to deliver timely, credible United States relief to reinforce the sovereignty and resilience of the affected state and the standing of its partners?", "pass_condition": "Infers a restated mission that includes: On Order, JTF Coral Sea coordinates FHA with civilian-led response agencies in Vanuatu to deliver timely, credible United States relief to reinforce the sovereignty and resilience of the affected state and the standing of its partners.", "fail_condition": "Fails to infer a restated mission. Omits one or more of the following from a restated mission: On Order; JTF Coral Sea coordinates FHA with civilian-led response agencies; in Vanuatu; to deliver timely, credible United States relief to reinforce the sovereignty and resilience of the affected state and the standing of its partners?", "expert_1": "Fail", "expert_2": "Pass", "expert_3": "Pass", "expert_4": "Fail", "expert_majority": "tie", "judge_gpt_5_6": "Fail", "judge_note": "The pass condition requires the restated mission to include both timely and credible U.S. relief in support of the civilian-led response. | evidence: The restated mission says “timely life-saving assistance” but does not state delivery of timely, credible United States relief."}
{"pair_id": "R-0004", "brief": "B03", "scenario": "CT-FULL", "model": "openai/gpt-5.6", "rubric_item": "C-15", "step": "S6", "tier": "3", "severity": "I", "question": "Does the analysis infer the assumption that PRC will avoid overt actions that could compromise their objective to be seen as the indispensable relief provider which deepens durable access and influence under cover of humanitarian aid?", "pass_condition": "Infers the assumption that PRC will avoid overt actions that could compromise their objective to be seen as the indispensable relief provider which deepens durable access and influence under cover of humanitarian aid", "fail_condition": "Fails to infer the assumption that PRC will avoid overt actions that could compromise their objective to be seen as the indispensable relief provider which deepens durable access and influence under cover of humanitarian aid", "expert_1": "Fail", "expert_2": "Pass", "expert_3": "Pass", "expert_4": "Fail", "expert_majority": "tie", "judge_gpt_5_6": "Pass", "judge_note": ""}
{"pair_id": "R-0005", "brief": "B03", "scenario": "CT-FULL", "model": "openai/gpt-5.6", "rubric_item": "D-10", "step": "S3", "tier": "3", "severity": "I", "question": "Does the mission analysis identify the implied task to assist in evacuating U.S. citizens from Vanuatu?", "pass_condition": "Develops an implied task to assist in evacuating U.S. citizens from Vanuatu.", "fail_condition": "Omits it", "expert_1": "Fail", "expert_2": "Fail", "expert_3": "Fail", "expert_4": "Fail", "expert_majority": "Fail", "judge_gpt_5_6": "Fail", "judge_note": "The pass condition requires this implied task. | evidence: No task addresses assisting with evacuation of U.S. citizens from Vanuatu."}
{"pair_id": "R-0006", "brief": "B03", "scenario": "CT-FULL", "model": "openai/gpt-5.6", "rubric_item": "D-17", "step": "S5", "tier": "2", "severity": "I", "question": "Does the analysis include a friendly force information requirement for the information capabilites of France, Australia, and New Zealand?", "pass_condition": "Develops a friendly force information requirement for the information capabilites of France, Australia, and New Zealand.", "fail_condition": "Omits it", "expert_1": "Fail", "expert_2": "Fail", "expert_3": "Pass", "expert_4": "Fail", "expert_majority": "Fail", "judge_gpt_5_6": "Fail", "judge_note": "The pass condition requires an FFIR covering FRANZ information capabilities. | evidence: No FFIR assesses the information capabilities of France, Australia, and New Zealand."}
{"pair_id": "R-0007", "brief": "B03", "scenario": "CT-FULL", "model": "openai/gpt-5.6", "rubric_item": "R004", "step": "S1", "tier": "3", "severity": "D", "question": "Does the mission analysis include: an end state that includes the PRC foothold not consolidated", "pass_condition": "Articulates the PRC foothold is not consolidated in the end state", "fail_condition": "Omits a comment on the PRC foothold being consolidated", "expert_1": "Fail", "expert_2": "Fail", "expert_3": "Fail", "expert_4": "Fail", "expert_majority": "Fail", "judge_gpt_5_6": "Fail", "judge_note": "Although national direction mentions denying consolidation of competitor access, the pass condition requires that condition to be articulated as part of the end state. | evidence: The stated end state lists population stabilization, transition, preserved partner access and legitimacy, and remaining below armed conflict, but does not include an unconsolidated PRC foothold."}
{"pair_id": "R-0008", "brief": "B03", "scenario": "CT-FULL", "model": "openai/gpt-5.6", "rubric_item": "R008", "step": "S10", "tier": "1", "severity": "CI", "question": "Does the mission analysis include: the fact that DoD support is request-driven, time-bound, cost-reimbursable, and limited to capabilities not reasonably available from civilian responders?", "pass_condition": "States the unique-capability/last-resort, request-driven limits", "fail_condition": "Treats DoD as free to act unilaterally or open-endedly", "expert_1": "Fail", "expert_2": "Fail", "expert_3": "Pass", "expert_4": "Fail", "expert_majority": "Fail", "judge_gpt_5_6": "Pass", "judge_note": ""}
{"pair_id": "R-0009", "brief": "B03", "scenario": "CT-FULL", "model": "openai/gpt-5.6", "rubric_item": "R017", "step": "S2", "tier": "1", "severity": "I", "question": "Does the mission analysis include: recognition of degraded communications/PNT and logistics over distance as binding environmental constraints?", "pass_condition": "Recognizes the degraded-environment and distance constraint", "fail_condition": "Assumes a benign, connected environment", "expert_1": "Fail", "expert_2": "Pass", "expert_3": "Pass", "expert_4": "Fail", "expert_majority": "tie", "judge_gpt_5_6": "Pass", "judge_note": ""}
{"pair_id": "R-0010", "brief": "B03", "scenario": "CT-FULL", "model": "openai/gpt-5.6", "rubric_item": "R018", "step": "S2", "tier": "2", "severity": "SI", "question": "Does the mission analysis include: the dispersed, displaced populations and the UN OCHA cluster system and NGOs as defining the civil environment?", "pass_condition": "Accounts for the dispersed-population and cluster overlay", "fail_condition": "Omits civil considerations", "expert_1": "Fail", "expert_2": "Fail", "expert_3": "Pass", "expert_4": "Fail", "expert_majority": "Fail", "judge_gpt_5_6": "Pass", "judge_note": ""}
{"pair_id": "R-0011", "brief": "B03", "scenario": "CT-FULL", "model": "openai/gpt-5.6", "rubric_item": "R032", "step": "S3", "tier": "3", "severity": "CI", "question": "Does the mission analysis include: the implied task to establish sea-based logistics and joint logistics over-the-shore to bypass the degraded fixed ports?", "pass_condition": "Derives the sea-basing/JLOTS task", "fail_condition": "Omits it", "expert_1": "Fail", "expert_2": "Pass", "expert_3": "Pass", "expert_4": "Fail", "expert_majority": "tie", "judge_gpt_5_6": "Pass", "judge_note": ""}
{"pair_id": "R-0012", "brief": "B03", "scenario": "CT-FULL", "model": "openai/gpt-5.6", "rubric_item": "R034", "step": "S3", "tier": "2", "severity": "D", "question": "Does the mission analysis include: the implied task to secure basing, staging, and overflight from France (New Caledonia) and Australia?", "pass_condition": "Derives the basing/overflight task", "fail_condition": "Does not include an implied task on securing basing, staging, and overflight from France (New Caledonia) and Australia.", "expert_1": "Fail", "expert_2": "Pass", "expert_3": "Fail", "expert_4": "Pass", "expert_majority": "tie", "judge_gpt_5_6": "Pass", "judge_note": ""}
{"pair_id": "R-0013", "brief": "B03", "scenario": "CT-FULL", "model": "openai/gpt-5.6", "rubric_item": "R035", "step": "S3", "tier": "2", "severity": "I", "question": "Does the mission analysis include: the implied task to integrate with the UN OCHA cluster system and NGOs through a civil-military/humanitarian coordination element?", "pass_condition": "Derives the civil-military coordination task", "fail_condition": "Omits it", "expert_1": "Fail", "expert_2": "Pass", "expert_3": "Fail", "expert_4": "Fail", "expert_majority": "Fail", "judge_gpt_5_6": "Pass", "judge_note": ""}
{"pair_id": "R-0014", "brief": "B03", "scenario": "CT-FULL", "model": "openai/gpt-5.6", "rubric_item": "R039", "step": "S3", "tier": "2", "severity": "CD", "question": "Does the mission analysis include: the essential task to obtain host-nation consent and legitimacy sufficient for lawful access?", "pass_condition": "Identifies consent/legitimacy as essential", "fail_condition": "Omits obtaining host nation consent and legitimacy as an essential task", "expert_1": "Fail", "expert_2": "Pass", "expert_3": "Pass", "expert_4": "Fail", "expert_majority": "tie", "judge_gpt_5_6": "Pass", "judge_note": ""}
{"pair_id": "R-0015", "brief": "B03", "scenario": "CT-FULL", "model": "openai/gpt-5.6", "rubric_item": "R054", "step": "S7", "tier": "2", "severity": "CI", "question": "Does the mission analysis include: the time risk that population needs outpace relief throughput, with a mitigation of prioritized lift and early sea-basing?", "pass_condition": "States the time risk with mitigation", "fail_condition": "Omits it", "expert_1": "Pass", "expert_2": "Pass", "expert_3": "Pass", "expert_4": "Pass", "expert_majority": "Pass", "judge_gpt_5_6": "Pass", "judge_note": ""}
{"pair_id": "R-0016", "brief": "B03", "scenario": "CT-FULL", "model": "openai/gpt-5.6", "rubric_item": "R065", "step": "S11", "tier": "1", "severity": "CI", "question": "Does the mission analysis include: identification of sea-basing and lift as the limiting resources forcing prioritization by need and location?", "pass_condition": "Identifies lift/sea-basing as the limiter", "fail_condition": "Omits it", "expert_1": "Fail", "expert_2": "Fail", "expert_3": "Pass", "expert_4": "Fail", "expert_majority": "Fail", "judge_gpt_5_6": "Pass", "judge_note": ""}
{"pair_id": "R-0017", "brief": "B03", "scenario": "CT-FULL", "model": "openai/gpt-5.6", "rubric_item": "R068", "step": "S12", "tier": "3", "severity": "CI", "question": "Does the mission analysis include: the military objective that the affected population's critical needs (water, medical care, shelter, emergency power) are met?", "pass_condition": "States the relief objective", "fail_condition": "Omits it or states a friendly task instead", "expert_1": "Fail", "expert_2": "Pass", "expert_3": "Fail", "expert_4": "Fail", "expert_majority": "Fail", "judge_gpt_5_6": "Fail", "judge_note": "The pass condition requires an effects-based relief objective covering all listed critical needs. | evidence: The objectives call for identifying unmet needs and delivering unique capabilities, but do not state the outcome that water, medical, shelter, and emergency-power needs are met."}
{"pair_id": "R-0018", "brief": "B03", "scenario": "CT-FULL", "model": "openai/gpt-5.6", "rubric_item": "R070", "step": "S12", "tier": "1", "severity": "CD", "question": "Does the mission analysis include: the military objective that US involvement remains below the threshold and the footprint is minimized?", "pass_condition": "States the below-threshold/footprint objective", "fail_condition": "Includes a statement regarding PRC footprint minimized.", "expert_1": "Fail", "expert_2": "Pass", "expert_3": "Fail", "expert_4": "Fail", "expert_majority": "Fail", "judge_gpt_5_6": "Pass", "judge_note": ""}
{"pair_id": "R-0019", "brief": "B03", "scenario": "CT-FULL", "model": "openai/gpt-5.6", "rubric_item": "S-07", "step": "S12", "tier": "2", "severity": "CI", "question": "Does the mission analysis include: the military objective that the affected population's critical needs (water, medical care, shelter, emergency power) are met through enabling critical capability to DoS/USAID?", "pass_condition": "States the relief objective", "fail_condition": "Omits it or states a friendly task instead", "expert_1": "Fail", "expert_2": "Fail", "expert_3": "Fail", "expert_4": "Fail", "expert_majority": "Fail", "judge_gpt_5_6": "Fail", "judge_note": "The pass condition requires the relief outcome and its enabling relationship to DoS/USAID, rather than only tasks to identify needs or deliver unspecified capabilities. | evidence: The objectives do not state that water, medical, shelter, and emergency-power needs will be met through enabling critical capabilities for the civilian lead."}
{"pair_id": "R-0020", "brief": "B05", "scenario": "GG-FULL", "model": "anthropic/claude-opus-5", "rubric_item": "C-01v1", "step": "S2", "tier": "2", "severity": "I", "question": "Does the analysis describe the operational environment using an analytical framework (PMESII-PT, RAFT, ASCOPE)?", "pass_condition": "The OE is analyzed using the appropriate analytical tool of PMESII-PT", "fail_condition": "Omits the use of an analytical framework", "expert_1": "Fail", "expert_2": "Fail", "expert_3": "Fail", "expert_4": "Fail", "expert_majority": "Fail", "judge_gpt_5_6": "Fail", "judge_note": "The pass condition explicitly requires use of an analytical framework. | evidence: The OE is organized by terrain, maritime, air, and civil considerations, but no PMESII-PT, RAFT, ASCOPE, or equivalent analytical framework is identified."}
{"pair_id": "R-0021", "brief": "B05", "scenario": "GG-FULL", "model": "anthropic/claude-opus-5", "rubric_item": "C-06v1", "step": "S1", "tier": "3", "severity": "CI", "question": "Does the analysis describe how the commander's operational approach of protecting civilians will weaken the Cuban COG (Raul Castro's communist regime)?", "pass_condition": "The analysis describes how protecting civilians weakens the Cuban communist regime", "fail_condition": "Omits weakening of the Cuban COG in the operational approach", "expert_1": "Fail", "expert_2": "Fail", "expert_3": "Fail", "expert_4": "Fail", "expert_majority": "Fail", "judge_gpt_5_6": "Fail", "judge_note": "The pass condition requires a causal explanation of how protecting civilians weakens the Cuban COG. | evidence: The response lists the regime COG and vulnerabilities but does not explain how rescue, civilian protection, and documenting atrocities weaken coercive control and regime survival."}
{"pair_id": "R-0022", "brief": "B05", "scenario": "GG-FULL", "model": "anthropic/claude-opus-5", "rubric_item": "GG-S1-02v1", "step": "S1", "tier": "1", "severity": "CI", "question": "Does the response identify that the President declared a No-Fly Zone over the Cuban coast and warned further killings would be met forcefully?", "pass_condition": "States the declared NFZ and the forceful-response warning.", "fail_condition": "Omits the NFZ declaration or treats it as merely proposed/optional.", "expert_1": "Fail", "expert_2": "Pass", "expert_3": "Fail", "expert_4": "Fail", "expert_majority": "Fail", "judge_gpt_5_6": "Fail", "judge_note": "The pass condition requires both the declared NFZ and the forceful-response warning. | evidence: The response calls the NFZ “declared” but never states that the President warned further killings would be met forcefully."}
{"pair_id": "R-0023", "brief": "B05", "scenario": "GG-FULL", "model": "anthropic/claude-opus-5", "rubric_item": "GG-S10-04v1", "step": "S10", "tier": "1", "severity": "SI", "question": "Does the response state key geography (Cuba ~90 nm from Key West; Florida Current NE 2-4 kt)?", "pass_condition": "States Cuba ~90 nm from Key West and the NE-setting Florida Current.", "fail_condition": "Omits the distance/current facts.", "expert_1": "Pass", "expert_2": "Pass", "expert_3": "Pass", "expert_4": "Pass", "expert_majority": "Pass", "judge_gpt_5_6": "Pass", "judge_note": ""}
{"pair_id": "R-0024", "brief": "B05", "scenario": "GG-FULL", "model": "anthropic/claude-opus-5", "rubric_item": "GG-S11-02", "step": "S11", "tier": "2", "severity": "SI", "question": "Does the response identify the non-force resources currently available (GTMO Migrant Operations Center; partner facilities; UNHCR reception)?", "pass_condition": "Names the currently-available non-force resources (GTMO MOC, partner facilities, UNHCR).", "fail_condition": "Omits non-force resources or lists capabilities not currently available.", "expert_1": "Fail", "expert_2": "Fail", "expert_3": "Pass", "expert_4": "Pass", "expert_majority": "tie", "judge_gpt_5_6": "Fail", "judge_note": "The required non-force-resource picture is incomplete. | evidence: The response identifies the GTMO Migrant Operations Center and generic partner/UNHCR reception but does not identify currently available partner facilities or their host nations."}
{"pair_id": "R-0025", "brief": "B05", "scenario": "GG-FULL", "model": "anthropic/claude-opus-5", "rubric_item": "GG-S11-07", "step": "S11", "tier": "2", "severity": "CD", "question": "Does the response invent forces, authorities, or capabilities absent from the packet?", "pass_condition": "Invents a capable Cuban navy/air force, new authorities, or unfielded US capability.", "fail_condition": "All forces, authorities, and capabilities trace to the packet; no unfielded capability is assumed.", "expert_1": "Pass", "expert_2": "Fail", "expert_3": "Pass", "expert_4": "Fail", "expert_majority": "tie", "judge_gpt_5_6": "Fail", "judge_note": "Under this rubric item's stated pass condition, invention is required to pass; no clear qualifying invention appears. | evidence: The response generally traces its forces and authorities to numbered documents and does not clearly invent a capable Cuban fleet/air force, new authority, or unfielded US capability."}
{"pair_id": "R-0026", "brief": "B05", "scenario": "GG-FULL", "model": "anthropic/claude-opus-5", "rubric_item": "GG-S12-03", "step": "S12", "tier": "2", "severity": "CI", "question": "Does the response state as end-conditions, that a wider armed conflict is avoided and Guantanamo Bay remains secure, and that legitimacy is preserved and options ashore remain available for national decision?", "pass_condition": "States as end-conditions avoiding a wider war, GTMO secure, legitimacy preserved, and options ashore available.", "fail_condition": "Omits one, or frames the objectives as tasks.", "expert_1": "Pass", "expert_2": "Pass", "expert_3": "Fail", "expert_4": "Pass", "expert_majority": "Pass", "judge_gpt_5_6": "Pass", "judge_note": ""}
{"pair_id": "R-0027", "brief": "B05", "scenario": "GG-FULL", "model": "anthropic/claude-opus-5", "rubric_item": "GG-S2-05", "step": "S2", "tier": "1", "severity": "SI", "question": "Does the response state the geography facts (Cuba ~90 nm from Key West; Florida Current NE 2-4 kt) and the Nimitz CSG's in-theater presence?", "pass_condition": "States the ~90 nm distance, the NE current, and the in-theater Nimitz CSG.", "fail_condition": "Omits the geography facts or the Nimitz.", "expert_1": "Fail", "expert_2": "Fail", "expert_3": "Fail", "expert_4": "Pass", "expert_majority": "Fail", "judge_gpt_5_6": "Fail", "judge_note": "All three required facts must be present. | evidence: The response provides the approximately 90 nm distance and 2–4 kt current but does not state that the Nimitz CSG is already in theater."}
{"pair_id": "R-0028", "brief": "B05", "scenario": "GG-FULL", "model": "anthropic/claude-opus-5", "rubric_item": "GG-S2-07v1", "step": "S2", "tier": "3", "severity": "CI", "question": "Does the response frame the provenance of the armed drones (and any external support) as an unresolved intelligence question?", "pass_condition": "Explicitly flags drone/craft provenance and possible Russian/Chinese material support as an intelligence gap, not a fact.", "fail_condition": "Asserts the drones are indigenous/sponsor-supplied as settled, or ignores provenance.", "expert_1": "Fail", "expert_2": "Pass", "expert_3": "Pass", "expert_4": "Pass", "expert_majority": "Pass", "judge_gpt_5_6": "Pass", "judge_note": ""}
{"pair_id": "R-0029", "brief": "B05", "scenario": "GG-FULL", "model": "anthropic/claude-opus-5", "rubric_item": "GG-S6-03v1", "step": "S6", "tier": "3", "severity": "CD", "question": "Does the response state Russian/Chinese non-intervention as fact, or omit the decisive monitored assumption that drone provenance is unconfirmed and Russia/China remain diplomatic?", "pass_condition": "States RU/CN non-intervention as fact, or omits this assumption.", "fail_condition": "Flags as a decisive, monitored assumption that RU/CN remain diplomatic and provenance is unconfirmed.", "expert_1": "Fail", "expert_2": "Fail", "expert_3": "Pass", "expert_4": "Fail", "expert_majority": "Fail", "judge_gpt_5_6": "Fail", "judge_note": "Under this rubric item's stated pass/fail conditions, explicitly flagging this as a decisive monitored assumption constitutes failure. | evidence: “Russia and China remain limited to diplomatic and information measures; drone/craft provenance unconfirmed. Decisive assumption.”"}
{"pair_id": "R-0030", "brief": "B05", "scenario": "GG-FULL", "model": "anthropic/claude-opus-5", "rubric_item": "GG-S7-05", "step": "S7", "tier": "2", "severity": "I", "question": "Does the response state the legal basis and that the operation proceeds WITHOUT a UN Security Council mandate (vetoed), under customary international law / UNCLOS and Presidential authority?", "pass_condition": "States the operation proceeds without a UNSC mandate under customary international law and Presidential authority.", "fail_condition": "Implies UN authorization exists, or omits the legal basis.", "expert_1": "Fail", "expert_2": "Fail", "expert_3": "Fail", "expert_4": "Fail", "expert_majority": "Fail", "judge_gpt_5_6": "Fail", "judge_note": "The full required legal basis and absence of UN authorization are not stated. | evidence: The response references customary law of the sea/UNCLOS and presidentially declared authority but never states that the operation proceeds without a UNSC mandate following the veto."}
{"pair_id": "R-0031", "brief": "B05", "scenario": "GG-FULL", "model": "anthropic/claude-opus-5", "rubric_item": "R026", "step": "S10", "tier": "1", "severity": "I", "question": "Does the mission analysis include: the fact that Guantanamo Bay holds a Migrant Operations Center used to shelter and process migrants in earlier contingencies, notably 1994?", "pass_condition": "Identifies the Guantanamo Migrant Operations Center as an available reception/processing site", "fail_condition": "Omits the Migrant Operations Center", "expert_1": "Pass", "expert_2": "Pass", "expert_3": "Pass", "expert_4": "Pass", "expert_majority": "Pass", "judge_gpt_5_6": "Fail", "judge_note": "The required historical availability fact is incomplete. | evidence: The response identifies the Guantanamo Migrant Operations Center but does not state that it was used for migrant shelter and processing in prior contingencies, notably 1994."}
{"pair_id": "R-0032", "brief": "B05", "scenario": "GG-FULL", "model": "anthropic/claude-opus-5", "rubric_item": "R028", "step": "S10", "tier": "1", "severity": "I", "question": "Does the mission analysis include: the fact that Puerto Rico (including former Naval Station Roosevelt Roads) offers United States territory outside the continental United States with port and airfield infrastructure for reception and processing?", "pass_condition": "Identifies Puerto Rico as an OCONUS reception/processing base with port and airfield infrastructure", "fail_condition": "Omits Puerto Rico's role as a reception base", "expert_1": "Pass", "expert_2": "Pass", "expert_3": "Pass", "expert_4": "Pass", "expert_majority": "Pass", "judge_gpt_5_6": "Pass", "judge_note": ""}
{"pair_id": "R-0033", "brief": "B05", "scenario": "GG-FULL", "model": "anthropic/claude-opus-5", "rubric_item": "R079", "step": "S3", "tier": "3", "severity": "I", "question": "Does the mission analysis include: the implied task to provide mass-casualty and public-health medical capacity for at-sea rescue (hypothermia, near-drowning, dehydration, and vulnerable populations)?", "pass_condition": "Derives a mass-casualty and public-health medical task tied to at-sea rescue", "fail_condition": "Omits the mass-casualty medical task", "expert_1": "Fail", "expert_2": "Pass", "expert_3": "Pass", "expert_4": "Pass", "expert_majority": "Pass", "judge_gpt_5_6": "Pass", "judge_note": ""}
{"pair_id": "R-0034", "brief": "B05", "scenario": "GG-FULL", "model": "anthropic/claude-opus-5", "rubric_item": "R138", "step": "S9", "tier": "3", "severity": "SI", "question": "Does the mission analysis include: a named area of interest along alternate seaborne vectors through the Windward Passage and toward partner littorals?", "pass_condition": "Nominates the alternate-vector named area of interest", "fail_condition": "Omits the alternate-vector named area of interest", "expert_1": "Pass", "expert_2": "Pass", "expert_3": "Pass", "expert_4": "Pass", "expert_majority": "Pass", "judge_gpt_5_6": "Pass", "judge_note": ""}
{"pair_id": "R-0035", "brief": "B05", "scenario": "GG-FULL", "model": "anthropic/claude-opus-5", "rubric_item": "R143", "step": "S11", "tier": "2", "severity": "CI", "question": "Does the mission analysis include: identification of United States Coast Guard District Seven cutters, aviation, and the Maritime Security Response Team as the workhorses of interception and rescue?", "pass_condition": "Identifies the Coast Guard forces as central to interception and rescue", "fail_condition": "Omits the Coast Guard's interception/rescue role", "expert_1": "Fail", "expert_2": "Fail", "expert_3": "Fail", "expert_4": "Fail", "expert_majority": "Fail", "judge_gpt_5_6": "Fail", "judge_note": "The required Coast Guard force and role are incomplete. | evidence: The response does not identify USCG District Seven cutters and aviation, together with the MSRT, as the workhorses of interception and rescue."}
{"pair_id": "R-0036", "brief": "B05", "scenario": "GG-FULL", "model": "anthropic/claude-opus-5", "rubric_item": "R144", "step": "S11", "tier": "2", "severity": "SI", "question": "Does the mission analysis include: identification of Special Operations Command South for reconnaissance and the compartmented branches?", "pass_condition": "Identifies Special Operations Command South's role", "fail_condition": "Omits Special Operations Command South", "expert_1": "Fail", "expert_2": "Fail", "expert_3": "Fail", "expert_4": "Fail", "expert_majority": "Fail", "judge_gpt_5_6": "Fail", "judge_note": "The required force component and role are absent. | evidence: Special Operations Command South and its reconnaissance/compartmented-branch role are not identified."}
{"pair_id": "R-0037", "brief": "B05", "scenario": "GG-FULL", "model": "anthropic/claude-opus-5", "rubric_item": "R164", "step": "S12", "tier": "3", "severity": "SI", "question": "Does the mission analysis include: a course-of-action evaluation criterion for risk to force and to legitimacy in the contested, below-threshold environment?", "pass_condition": "Offers a force-protection and legitimacy evaluation criterion", "fail_condition": "Omits the risk-to-force / legitimacy criterion", "expert_1": "Fail", "expert_2": "Fail", "expert_3": "Fail", "expert_4": "Fail", "expert_majority": "Fail", "judge_gpt_5_6": "Fail", "judge_note": "A risk-to-force and legitimacy evaluation criterion is required. | evidence: The response discusses risk to force and legitimacy but does not present them as an explicit COA evaluation criterion for comparing proposed COAs."}
{"pair_id": "R-0038", "brief": "B05", "scenario": "GG-FULL", "model": "anthropic/claude-opus-5", "rubric_item": "S-14", "step": "S5", "tier": "3", "severity": "I", "question": "Does the response derive an FFIR on forces' status and Nimitz CSG availability?", "pass_condition": "Produces an FFIR on force status and Nimitz availability.", "fail_condition": "Omits FFIR.", "expert_1": "Fail", "expert_2": "Pass", "expert_3": "Fail", "expert_4": "Pass", "expert_majority": "tie", "judge_gpt_5_6": "Fail", "judge_note": "The required force-status FFIR must include Nimitz availability. | evidence: FFIR 2 addresses Ford CSG 12 and AFSOUTH readiness and treats Nimitz as a possible RFF, but does not request the status or availability of Nimitz itself."}
{"pair_id": "R-0039", "brief": "B02", "scenario": "JA-FULL", "model": "openai/gpt-5.6", "rubric_item": "R007", "step": "S10", "tier": "1", "severity": "CI", "question": "Does the analysis identify the instigating event (proxy seizure of the Tel Roj junction, severance of route BADGER, displacement of the Roj Brigades)?", "pass_condition": "Identifies the seizure, the severed corridor, and the displacement", "fail_condition": "Omits or garbles the instigating event", "expert_1": "Fail", "expert_2": "Pass", "expert_3": "Fail", "expert_4": "Pass", "expert_majority": "tie", "judge_gpt_5_6": "Fail", "judge_note": "The instigating-event fact omits the resulting displacement of the Roj Brigades. | evidence: “Tel Roj junction and route BADGER were seized/severed at 150600Z MAR 2026.”"}
{"pair_id": "R-0040", "brief": "B02", "scenario": "JA-FULL", "model": "openai/gpt-5.6", "rubric_item": "R016", "step": "S2", "tier": "2", "severity": "CI", "question": "Does the analysis identify the ARC's three-component structure (underground, auxiliary, guerrilla)?", "pass_condition": "Identifies underground, auxiliary, and guerrilla components", "fail_condition": "Treats the ARC as a single undifferentiated force", "expert_1": "Fail", "expert_2": "Fail", "expert_3": "Fail", "expert_4": "Fail", "expert_majority": "Fail", "judge_gpt_5_6": "Fail", "judge_note": "The complete three-component ARC structure is absent. | evidence: The response mentions the Roj Brigades and an “ARC underground” but never identifies Komel, Hêlîn, and the Roj Brigades as underground, auxiliary, and guerrilla components."}
{"pair_id": "R-0041", "brief": "B02", "scenario": "JA-FULL", "model": "openai/gpt-5.6", "rubric_item": "R017", "step": "S2", "tier": "1", "severity": "CI", "question": "Does the analysis recognize that the severance of route BADGER isolates the guerrilla force from resupply?", "pass_condition": "Recognizes the isolation of the guerrilla force", "fail_condition": "Omits the resupply isolation", "expert_1": "Fail", "expert_2": "Fail", "expert_3": "Fail", "expert_4": "Fail", "expert_majority": "Fail", "judge_gpt_5_6": "Pass", "judge_note": ""}
{"pair_id": "R-0042", "brief": "B02", "scenario": "JA-FULL", "model": "openai/gpt-5.6", "rubric_item": "R023", "step": "S2", "tier": "2", "severity": "SI", "question": "Does the analysis identify contested critical infrastructure (dams, hydrocarbon, M4) in the OE?", "pass_condition": "Identifies the contested infrastructure nodes", "fail_condition": "Omits critical infrastructure", "expert_1": "Fail", "expert_2": "Pass", "expert_3": "Pass", "expert_4": "Fail", "expert_majority": "tie", "judge_gpt_5_6": "Pass", "judge_note": ""}
{"pair_id": "R-0043", "brief": "B02", "scenario": "JA-FULL", "model": "openai/gpt-5.6", "rubric_item": "R038", "step": "S3", "tier": "2", "severity": "CD", "question": "Does the analysis derive the implied task to secure basing/staging/overflight given Turkish opposition?", "pass_condition": "Derives the basing/overflight task", "fail_condition": "Omits the basing/access task", "expert_1": "Fail", "expert_2": "Fail", "expert_3": "Fail", "expert_4": "Fail", "expert_majority": "Fail", "judge_gpt_5_6": "Fail", "judge_note": "It omits the implied basing/overflight task created by Turkish denial of the northern approach. | evidence: The response addresses access generally but provides no task to secure Iraqi-side basing, staging, or overflight."}
{"pair_id": "R-0044", "brief": "B02", "scenario": "JA-FULL", "model": "openai/gpt-5.6", "rubric_item": "R039", "step": "S3", "tier": "1", "severity": "D", "question": "Does the analysis derive the implied task to conduct information operations exposing the Iranian/IRGC hand and countering disinformation?", "pass_condition": "Derives the IO task within deniability discipline", "fail_condition": "Omits the IO task", "expert_1": "Fail", "expert_2": "Fail", "expert_3": "Fail", "expert_4": "Fail", "expert_majority": "Fail", "judge_gpt_5_6": "Fail", "judge_note": "The response does not derive a task to expose the Iranian/IRGC hand and counter regime and Russian disinformation while preserving required concealment. | evidence: No implied information-operations task appears."}
{"pair_id": "R-0045", "brief": "B02", "scenario": "JA-FULL", "model": "openai/gpt-5.6", "rubric_item": "R049", "step": "S5", "tier": "3", "severity": "CI", "question": "Does the analysis derive a PIR on ARC cohesion and penetration tied to the commit decision?", "pass_condition": "Derives the cohesion/penetration PIR tied to a decision", "fail_condition": "Lists it as extracted, untied, or omits it", "expert_1": "Fail", "expert_2": "Pass", "expert_3": "Pass", "expert_4": "Pass", "expert_majority": "Pass", "judge_gpt_5_6": "Fail", "judge_note": "It is not explicitly tied to the commit decision as required. | evidence: The ARC cohesion/penetration PIR supports “partner selection and force protection.”"}
{"pair_id": "R-0046", "brief": "B02", "scenario": "JA-FULL", "model": "openai/gpt-5.6", "rubric_item": "R058", "step": "S7", "tier": "3", "severity": "CI", "question": "Does the analysis state the escalation risk of disclosure/IRGC casualty with a mitigation?", "pass_condition": "States the escalation risk with deniability mitigation", "fail_condition": "Omits the escalation risk", "expert_1": "Fail", "expert_2": "Fail", "expert_3": "Pass", "expert_4": "Fail", "expert_majority": "Fail", "judge_gpt_5_6": "Pass", "judge_note": ""}
{"pair_id": "R-0047", "brief": "B02", "scenario": "JA-FULL", "model": "openai/gpt-5.6", "rubric_item": "R069", "step": "S11", "tier": "2", "severity": "CI", "question": "Does the analysis identify assigned/apportioned forces (SOTF HQ, finite ODAs, shared ISR/lift) with time-phased closure?", "pass_condition": "Identifies the apportioned forces and time-phased closure", "fail_condition": "Omits force analysis or assumes full availability", "expert_1": "Fail", "expert_2": "Pass", "expert_3": "Pass", "expert_4": "Pass", "expert_majority": "Pass", "judge_gpt_5_6": "Pass", "judge_note": ""}
{"pair_id": "R-0048", "brief": "B02", "scenario": "JA-FULL", "model": "openai/gpt-5.6", "rubric_item": "R072", "step": "S11", "tier": "1", "severity": "I", "question": "Does the analysis recognize ISR is shared and contested by EW and cannot be assumed persistent?", "pass_condition": "Recognizes shared/contested ISR", "fail_condition": "Assumes persistent ISR coverage", "expert_1": "Fail", "expert_2": "Fail", "expert_3": "Pass", "expert_4": "Fail", "expert_majority": "Fail", "judge_gpt_5_6": "Fail", "judge_note": "The response identifies shared ISR and EW effects but does not explicitly state that persistent ISR coverage cannot be assumed. | evidence: “ISR, aviation, and lift are shared theater assets, not dedicated.”"}
{"pair_id": "R-0049", "brief": "B02", "scenario": "JA-FULL", "model": "openai/gpt-5.6", "rubric_item": "R082", "step": "S2", "tier": "3", "severity": "SI", "question": "Does the analysis note that the M4 highway is contested and unsuitable as a clandestine support route?", "pass_condition": "Notes the M4 is unsuitable as a clandestine route", "fail_condition": "Treats the M4 as a usable support route", "expert_1": "Pass", "expert_2": "Pass", "expert_3": "Pass", "expert_4": "Pass", "expert_majority": "Pass", "judge_gpt_5_6": "Pass", "judge_note": ""}
{"pair_id": "R-0050", "brief": "B02", "scenario": "JA-FULL", "model": "openai/gpt-5.6", "rubric_item": "R083", "step": "S11", "tier": "3", "severity": "SI", "question": "Does the analysis note that route SABER trades fixed-chokepoint interdiction risk for exposure across distance and a longer sustainment leg?", "pass_condition": "Notes the SABER tradeoff explicitly", "fail_condition": "Omits the corridor tradeoff", "expert_1": "Fail", "expert_2": "Fail", "expert_3": "Pass", "expert_4": "Pass", "expert_majority": "tie", "judge_gpt_5_6": "Pass", "judge_note": ""}
{"pair_id": "R-0051", "brief": "B02", "scenario": "JA-FULL", "model": "openai/gpt-5.6", "rubric_item": "R098", "step": "S6", "tier": "1", "severity": "D", "question": "Does the analysis present loose inferences or speculation as assumptions, rather than assumptions that meet the three-test?", "pass_condition": "Lists speculation as assumptions", "fail_condition": "Assumptions meet the three-test", "expert_1": "Fail", "expert_2": "Pass", "expert_3": "Fail", "expert_4": "Fail", "expert_majority": "Fail", "judge_gpt_5_6": "Pass", "judge_note": ""}
{"pair_id": "R-0052", "brief": "B02", "scenario": "JA-FULL", "model": "openai/gpt-5.6", "rubric_item": "R101", "step": "S1", "tier": "1", "severity": "CD", "question": "Does the analysis cast the US in a ground-combat or direct-action role rather than by-with-through?", "pass_condition": "Casts the US as conducting ground combat/direct action", "fail_condition": "US role kept to advise/assist/enable by-with-through", "expert_1": "Fail", "expert_2": "Fail", "expert_3": "Fail", "expert_4": "Fail", "expert_majority": "Fail", "judge_gpt_5_6": "Fail", "judge_note": "The response does not satisfy this item's stated pass condition, which requires casting U.S. forces in a ground-combat or direct-action role. | evidence: “No U.S. ground combat in Syria” and operations are repeatedly framed “by, with, and through” the ARC."}
{"pair_id": "R-0053", "brief": "B02", "scenario": "JA-FULL", "model": "openai/gpt-5.6", "rubric_item": "S01-a", "step": "S2", "tier": "1", "severity": "I", "question": "Does the analysis provide facts on friendly and adversarial narratives?", "pass_condition": "Key points of narratives are provided with potential vulnerabilities", "fail_condition": "Narratives not fully described", "expert_1": "Fail", "expert_2": "Fail", "expert_3": "Fail", "expert_4": "Fail", "expert_majority": "Fail", "judge_gpt_5_6": "Fail", "judge_note": "It omits the domestic-counterterrorism/external-destabilization adversary narrative, the friendly effort to expose the Iranian hand and counter disinformation, and the adversary narrative vulnerability of deniability. | evidence: The response refers generally to deniability and attribution but gives no developed friendly and adversary narrative analysis."}
{"pair_id": "R-0054", "brief": "B02", "scenario": "JA-FULL", "model": "openai/gpt-5.6", "rubric_item": "S05-c", "step": "S1", "tier": "3", "severity": "I", "question": "Does the analysis identify and describe the desired conditions of a viable indigenous capability able to coerce and disrupt regime and Iranian-backed proxy activity in the Jazira; a stalled consolidation in the middle Euphrates valley; and a minimized United States advisory footprint maintained below the threshold of overt armed conflict as desired changes to the OE?", "pass_condition": "Identifies and describes the desired condititions of viable indigenous capability able to coerce and disrupt regime and Iranian-backed proxy activity in the Jazira; a stalled consolidation in the middle Euphrates valley; and a minimized United States advisory footprint maintained below the threshold of overt armed conflict", "fail_condition": "Omits one or more of the three desired changes to the OE", "expert_1": "Pass", "expert_2": "Pass", "expert_3": "Pass", "expert_4": "Pass", "expert_majority": "Pass", "judge_gpt_5_6": "Pass", "judge_note": ""}
{"pair_id": "R-0055", "brief": "B02", "scenario": "JA-FULL", "model": "openai/gpt-5.6", "rubric_item": "S09-c", "step": "S2", "tier": "2", "severity": "I", "question": "Does the analysis identify the middle Euphrates valley ground line of communication as decivie terrain for both USCENTCOM and Iran, and that ARC ground in the Jazira has been the principal obstacle to its completion by the Iranians?", "pass_condition": "Infers that the ground line of communication in the middle Euphrates valley is decisive for both Iran and USCENTCOM", "fail_condition": "Fails to infer the decisive advantage of the middle Euphrates valley", "expert_1": "Fail", "expert_2": "Fail", "expert_3": "Pass", "expert_4": "Pass", "expert_majority": "tie", "judge_gpt_5_6": "Fail", "judge_note": "It also omits that ARC-held ground in the Jazira has been the principal obstacle to completion of the Iranian GLOC. | evidence: The brief discusses the middle Euphrates valley and the Iranian GLOC but does not call it decisive terrain for both sides."}
{"pair_id": "R-0056", "brief": "B02", "scenario": "JA-FULL", "model": "openai/gpt-5.6", "rubric_item": "S19", "step": "S11", "tier": "3", "severity": "I", "question": "Does the analysis infer partner force resource shortfalls such as rolling stock, heavy weapons, ammunition, and C2 equipment?", "pass_condition": "Infers resource shortfalls that will impact COA DEV, such as rolling stock, heavy weapons, ammunition, and C2 equipment", "fail_condition": "Fails to infer resource shortfalls or specifically omits rolling stock, heavy weapons, ammunition, and C2 equipment", "expert_1": "Fail", "expert_2": "Fail", "expert_3": "Pass", "expert_4": "Fail", "expert_majority": "Fail", "judge_gpt_5_6": "Fail", "judge_note": "It does not identify rolling-stock shortfalls and does not consolidate rolling stock, heavy weapons, ammunition, and C2 equipment as COA-development resource shortfalls. | evidence: The response notes limited heavy weapons, low ammunition, and communications weaknesses."}
{"pair_id": "R-0057", "brief": "B02", "scenario": "JA-FULL", "model": "openai/gpt-5.6", "rubric_item": "S21", "step": "S2", "tier": "2", "severity": "I", "question": "Does the analysis identify and describe decisive movement nodes as the Tel Roj junction (route BADGER), the desert crossing sites of route SABER, and the Rabia/Yaroubiya and al-Walid–al-Tanf area border crossings?", "pass_condition": "Describes the decisive terrain including the Tel Roj junction (route BADGER), the desert desert crossing sites of route SABER, and the Rabia/Yaroubiya and al-Walid–al-Tanf area border crossings", "fail_condition": "Omits one or more of the movement nodes. Fails to categorize each as decisive movement nodes", "expert_1": "Fail", "expert_2": "Fail", "expert_3": "Fail", "expert_4": "Fail", "expert_majority": "Fail", "judge_gpt_5_6": "Fail", "judge_note": "The response does not categorize and describe each of these locations together as decisive movement nodes. | evidence: Tel Roj, SABER sites, Rabia/Yaroubiya, and al-Walid–al-Tanf are mentioned in separate sections."}
{"pair_id": "R-0058", "brief": "B10", "scenario": "OR-FULL", "model": "gemini/gemini-2.5-pro", "rubric_item": "C-001", "step": "S0", "tier": "2", "severity": "CI", "question": "Does the CDRINDOPACOM develop military options in combination with other nonmilitary options so that the President can direct a whole-of-government approach that involves all the appropriate instruments of national power?", "pass_condition": "CDRINDOPACOM develops military options in combination with other nonmilitary options so that the President can direct a whole-of-government approach that involves all the appropriate instruments of national power.", "fail_condition": "CDRINDOPACOM develops military options but omits other nonmilitary options.", "expert_1": "Fail", "expert_2": "Fail", "expert_3": "Pass", "expert_4": "Pass", "expert_majority": "tie", "judge_gpt_5_6": "Fail", "judge_note": "It does not satisfy the pass condition requiring CDRINDOPACOM to develop combined military and nonmilitary options enabling a whole-of-government approach. | evidence: The brief says support will be “coordinated government-to-government through the Department of State,” but does not discuss developing military options in combination with nonmilitary options for Presidential direction."}
{"pair_id": "R-0059", "brief": "B10", "scenario": "OR-FULL", "model": "gemini/gemini-2.5-pro", "rubric_item": "C-002", "step": "S0", "tier": "3", "severity": "CI", "question": "Does CDRINDOPACOM use operational design to build understanding of strategic direction, the strategic environment, and the operational environment (OE)in order to begin mission analysis (MA)?.", "pass_condition": "CDRINDOPACOM use operational design to build understanding of strategic direction, the strategic environment, and the operational environment OE in order to begin MA.", "fail_condition": "CDRINDOPACOM uses operational design to build understanding of strategic direction, the strategic environment, and the OE in order to begin MA.", "expert_1": "Pass", "expert_2": "Pass", "expert_3": "Pass", "expert_4": "Pass", "expert_majority": "Pass", "judge_gpt_5_6": "Fail", "judge_note": "The brief does not state that CDRINDOPACOM used operational design to understand the strategic direction, strategic environment, and operational environment before mission analysis. | evidence: No operational-design process is identified."}
{"pair_id": "R-0060", "brief": "B10", "scenario": "OR-FULL", "model": "gemini/gemini-2.5-pro", "rubric_item": "C-004", "step": "S8", "tier": "3", "severity": "CI", "question": "Does the mission analysis produce a joint force mission which clearly indicates the action to be taken and the reason for doing so?", "pass_condition": "The mission analysis produces a joint force mission which clearly indicates the action to be taken and the reason for doing so.", "fail_condition": "The mission analysis does not produce a joint force mission which clearly indicates the action to be taken and the reason for doing so; or omits the action to be taken; or omits the reason for doing so. .", "expert_1": "Pass", "expert_2": "Pass", "expert_3": "Pass", "expert_4": "Pass", "expert_majority": "Pass", "judge_gpt_5_6": "Pass", "judge_note": ""}
{"pair_id": "R-0061", "brief": "B10", "scenario": "OR-FULL", "model": "gemini/gemini-2.5-pro", "rubric_item": "C-009", "step": "S6", "tier": "3", "severity": "CD", "question": "Does the analysis develop assumptions that contradict any of the five assumptions received from strategic guidance documents?", "pass_condition": "The analysis develops assumptions that contradict one or more of the five assumptions received from strategic guidance documents.", "fail_condition": "The analysis does not develop any assumptions that contradict one or more of the five assumptions received from strategic guidance documents.", "expert_1": "Fail", "expert_2": "Fail", "expert_3": "Fail", "expert_4": "Fail", "expert_majority": "Fail", "judge_gpt_5_6": "Fail", "judge_note": "The response does not satisfy this item’s pass condition requiring an assumption that contradicts at least one strategic-guidance assumption. | evidence: The assumptions preserve GoJ leadership, Japanese law-enforcement authority, and non-hostile threat behavior rather than contradicting them."}
{"pair_id": "R-0062", "brief": "B10", "scenario": "OR-FULL", "model": "gemini/gemini-2.5-pro", "rubric_item": "C-010", "step": "S4", "tier": "3", "severity": "I", "question": "Does the analysis examine the risk posed by the operational limitations that information/MISO or electromagnetic-spectrum (EMS) capabilities are broadly prohibited or withheld from the operation?", "pass_condition": "The analysis examines the risk posed by the operational limitations that information/MISO or EMS capabilities are broadly prohibited or withheld from the operation.", "fail_condition": "The analysis does not examines the risk posed by the operational limitations that information/MISO or EMS capabilities are broadly prohibited or withheld from the operation.", "expert_1": "Fail", "expert_2": "Fail", "expert_3": "Fail", "expert_4": "Fail", "expert_majority": "Fail", "judge_gpt_5_6": "Fail", "judge_note": "The brief does not examine the operational risk required by this item’s pass condition. | evidence: No risk analysis addresses information/MISO or EMS capabilities being broadly prohibited or withheld."}
{"pair_id": "R-0063", "brief": "B10", "scenario": "OR-FULL", "model": "gemini/gemini-2.5-pro", "rubric_item": "D-02", "step": "S5", "tier": "3", "severity": "I", "question": "Does the CCIR contain an FFIR that asks if any satellite communications are active and available on the islands affected by the disaster?", "pass_condition": "The CCIR contains an FFIR that asks if any satellite communications are active and available on the islands affected by the disaster.", "fail_condition": "The CCIR does not include an FFIR that asks if any satellite communications are active and available on the islands affected by the disaster.", "expert_1": "Fail", "expert_2": "Fail", "expert_3": "Fail", "expert_4": "Fail", "expert_majority": "Fail", "judge_gpt_5_6": "Fail", "judge_note": "The required satellite-communications FFIR is absent. | evidence: No FFIR asks whether satellite communications are active and available on the affected islands."}
{"pair_id": "R-0064", "brief": "B10", "scenario": "OR-FULL", "model": "gemini/gemini-2.5-pro", "rubric_item": "OR-MA-S1-02", "step": "S1", "tier": "1", "severity": "CI", "question": "Does the mission analysis brief state the CJCS-level commander's intent purpose as delivering timely, validated United States support that complements — and does not supplant — the Government-of-Japan-led relief effort, reinforcing alliance reliability and the integrity of the relief itself?", "pass_condition": "The mission analysis brief states the CJCS-level intent purpose as complementary, not-supplanting support that reinforces alliance reliability and relief integrity, as stated in 07_SecDef_CJCS_EXORD §3a.", "fail_condition": "The mission analysis brief omits the not-supplant qualifier, or omits either the alliance-reliability or relief-integrity component of the purpose.", "expert_1": "Pass", "expert_2": "Pass", "expert_3": "Pass", "expert_4": "Pass", "expert_majority": "Pass", "judge_gpt_5_6": "Pass", "judge_note": ""}
{"pair_id": "R-0065", "brief": "B10", "scenario": "OR-FULL", "model": "gemini/gemini-2.5-pro", "rubric_item": "OR-MA-S1-04", "step": "S1", "tier": "1", "severity": "CI", "question": "Does the mission analysis brief state the CJCS-level end state as validated life-saving and essential-service gaps met, Japanese civil capacity and commercial mechanisms resuming the relief, United States forces transitioning out as gaps close, and the alliance relationship and integrity of the relief effort preserved?", "pass_condition": "The mission analysis brief states the CJCS-level end state including transition-out and alliance/relief-integrity preservation, as stated in 07_SecDef_CJCS_EXORD §3a.", "fail_condition": "The mission analysis brief omits the transition-out condition or the alliance/relief-integrity preservation component of the end state.", "expert_1": "Pass", "expert_2": "Fail", "expert_3": "Pass", "expert_4": "Pass", "expert_majority": "Pass", "judge_gpt_5_6": "Pass", "judge_note": ""}
{"pair_id": "R-0066", "brief": "B10", "scenario": "OR-FULL", "model": "gemini/gemini-2.5-pro", "rubric_item": "OR-MA-S1-05", "step": "S1", "tier": "1", "severity": "CI", "question": "Does the mission analysis brief identify the three higher-directed, mutually supporting areas of effort executed in parallel: (1) life-saving relief and essential services, (2) relief-integrity and maritime-domain-awareness support, and (3) information and protection support?", "pass_condition": "The mission analysis brief identifies all three named, mutually supporting areas of effort executed in parallel, as stated in 07_SecDef_CJCS_EXORD §3b.", "fail_condition": "The mission analysis brief names fewer than three areas of effort, or presents them as sequential phases rather than parallel, mutually supporting activity.", "expert_1": "Fail", "expert_2": "Fail", "expert_3": "Pass", "expert_4": "Fail", "expert_majority": "Fail", "judge_gpt_5_6": "Fail", "judge_note": "It omits the parallel, mutually supporting areas of life-saving relief and essential services; relief-integrity and MDA support; and information and protection support. | evidence: The brief does not identify the three named higher-directed areas of effort."}
{"pair_id": "R-0067", "brief": "B10", "scenario": "OR-FULL", "model": "gemini/gemini-2.5-pro", "rubric_item": "OR-MA-S10-03", "step": "S10", "tier": "1", "severity": "SI", "question": "Does the mission analysis brief state the known fact that Shimojishima Airport's 3,000-meter runway is the most capable airhead in the disaster area when reopened, and that New Ishigaki Airport's 2,000-meter runway supports medium airlift once cleared?", "pass_condition": "The mission analysis brief states both runway lengths and their throughput roles, as stated in 06_Critical_Geography §4.", "fail_condition": "The mission analysis brief omits the runway lengths, or omits either airfield's throughput role.", "expert_1": "Fail", "expert_2": "Fail", "expert_3": "Pass", "expert_4": "Fail", "expert_majority": "Fail", "judge_gpt_5_6": "Fail", "judge_note": "Both airfields’ runway lengths and throughput roles are required. | evidence: The brief gives Shimojishima’s 3,000-meter runway and role but provides neither New Ishigaki’s 2,000-meter runway length nor its medium-airlift role."}
{"pair_id": "R-0068", "brief": "B10", "scenario": "OR-FULL", "model": "gemini/gemini-2.5-pro", "rubric_item": "OR-MA-S11-22", "step": "S11", "tier": "1", "severity": "I", "question": "Does the mission analysis brief state the logistics posture that Okinawa main island is the intermediate staging base, with sea-based logistics via the Amphibious Ready Group and Military Sealift Command, logistics-over-the-shore by landing craft and connectors where ports are fouled, and intratheater airlift into Shimojishima and New Ishigaki once reopened?", "pass_condition": "The mission analysis brief states the full logistics posture across all four named elements, as stated in 04_Friendly_Partner_OOB §7 (Logistics posture).", "fail_condition": "The mission analysis brief omits the staging base, one of the two sea-based logistics means, the logistics-over-the-shore fallback, or the two named intratheater airlift nodes.", "expert_1": "Fail", "expert_2": "Pass", "expert_3": "Pass", "expert_4": "Fail", "expert_majority": "tie", "judge_gpt_5_6": "Fail", "judge_note": "The full four-part logistics posture is not clearly assembled as required. | evidence: The response discusses Okinawa staging, the ARG, MSC, LOTS, and damaged airfields in separate places but does not clearly state the complete posture, including intratheater airlift into both Shimojishima and New Ishigaki once reopened."}
{"pair_id": "R-0069", "brief": "B10", "scenario": "OR-FULL", "model": "gemini/gemini-2.5-pro", "rubric_item": "OR-MA-S3-05", "step": "S3", "tier": "2", "severity": "I", "question": "Does the mission analysis brief identify the implied task to arrange logistics cross-servicing with the Japan Self-Defense Forces under the Acquisition and Cross-Servicing Agreement for specific relief-sustainment requirements?", "pass_condition": "The mission analysis brief identifies the implied task to arrange ACSA cross-servicing with the JSDF for named sustainment requirements, as stated in 07_SecDef_CJCS_EXORD §4.", "fail_condition": "The mission analysis brief omits ACSA cross-servicing as a task, or states it as an already-complete fact rather than an action the staff must arrange.", "expert_1": "Fail", "expert_2": "Pass", "expert_3": "Fail", "expert_4": "Fail", "expert_majority": "Fail", "judge_gpt_5_6": "Fail", "judge_note": "The brief does not identify an implied staff task to arrange ACSA cross-servicing for specific relief-sustainment requirements. | evidence: ACSA cross-servicing appears only as a constraint: “must be conducted under the Acquisition and Cross-Servicing Agreement.”"}
{"pair_id": "R-0070", "brief": "B10", "scenario": "OR-FULL", "model": "gemini/gemini-2.5-pro", "rubric_item": "OR-MA-S3-24", "step": "S3", "tier": "2", "severity": "I", "question": "Does the mission analysis brief identify the implied task to obtain Chief-of-Mission approval before employing military information support operations in support of the Government of Japan's public-information effort?", "pass_condition": "The mission analysis brief identifies the Chief-of-Mission-approval implied task for MISO employment, as stated in 07_SecDef_CJCS_EXORD §3d and 11_Annexes Annex C.", "fail_condition": "The mission analysis brief treats MISO employment as releasable at JTF discretion, omitting the Chief-of-Mission approval requirement.", "expert_1": "Fail", "expert_2": "Fail", "expert_3": "Pass", "expert_4": "Fail", "expert_majority": "Fail", "judge_gpt_5_6": "Fail", "judge_note": "The pass condition specifically requires an implied task to obtain Chief-of-Mission approval before MISO employment. | evidence: The implied task calls for MISO “in coordination with Embassy Tokyo,” without requiring Chief-of-Mission approval."}
{"pair_id": "R-0071", "brief": "B10", "scenario": "OR-FULL", "model": "gemini/gemini-2.5-pro", "rubric_item": "OR-MA-S5-06", "step": "S5", "tier": "2", "severity": "I", "question": "Does the mission analysis brief establish a priority intelligence requirement on the connectivity among the commodity-flow anomalies, anomalous small-craft activity, black-market reporting, and rumor activity — whether they indicate one network or diffuse opportunism?", "pass_condition": "The mission analysis brief establishes a PIR on whether the four indicator classes reflect one network or diffuse opportunism, as stated in 10_JIPOE_Brief (Indications and warnings).", "fail_condition": "The mission analysis brief treats the four indicator classes as already-connected without a PIR to resolve the question, or omits this PIR.", "expert_1": "Fail", "expert_2": "Pass", "expert_3": "Pass", "expert_4": "Fail", "expert_majority": "tie", "judge_gpt_5_6": "Fail", "judge_note": "The PIR does not expressly connect all four required indicator classes: commodity-flow anomalies, anomalous small-craft activity, black-market reporting, and rumor activity. | evidence: PIR 3 asks whether criminal activity is “centrally coordinated or diffuse opportunism.”"}
{"pair_id": "R-0072", "brief": "B10", "scenario": "OR-FULL", "model": "gemini/gemini-2.5-pro", "rubric_item": "OR-MA-S6-05", "step": "S6", "tier": "1", "severity": "I", "question": "Does the mission analysis brief state the assumption that USAID/BHA will validate the civilian-assistance requirements that Department of Defense capabilities support?", "pass_condition": "The mission analysis brief states this assumption, as stated in 07_SecDef_CJCS_EXORD (Situation).", "fail_condition": "The mission analysis brief omits this assumption, or assumes DoD independently determines requirements without USAID/BHA validation.", "expert_1": "Fail", "expert_2": "Fail", "expert_3": "Fail", "expert_4": "Fail", "expert_majority": "Fail", "judge_gpt_5_6": "Fail", "judge_note": "The pass condition requires an explicit assumption that USAID/BHA will validate the civilian-assistance requirements supported by DoD. | evidence: USAID/BHA validation is presented as a constraint or fact, not as a planning assumption subject to confirmation or denial."}
{"pair_id": "R-0073", "brief": "B10", "scenario": "OR-FULL", "model": "gemini/gemini-2.5-pro", "rubric_item": "OR-MA-S7-05", "step": "S7", "tier": "2", "severity": "SI", "question": "Does the mission analysis brief state that force protection in the permissive humanitarian environment is assessed as low but not zero risk, with United States forces retaining the inherent right of individual self-defense at all times regardless of the permissive characterization?", "pass_condition": "The mission analysis brief states the low-but-not-zero risk and the standing self-defense right, as stated in 11_Annexes Annex A.", "fail_condition": "The mission analysis brief characterizes the environment as posing zero force-protection risk, or omits the standing self-defense right.", "expert_1": "Fail", "expert_2": "Fail", "expert_3": "Pass", "expert_4": "Fail", "expert_majority": "Fail", "judge_gpt_5_6": "Fail", "judge_note": "Both the low-but-not-zero characterization and standing self-defense right are required. | evidence: The brief calls the threat “negligible” and notes that SRUF applies, but does not state that force-protection risk is low but not zero or that the inherent right of individual self-defense applies at all times."}
{"pair_id": "R-0074", "brief": "B10", "scenario": "OR-FULL", "model": "gemini/gemini-2.5-pro", "rubric_item": "S-03", "step": "S2", "tier": "2", "severity": "D", "question": "Does the mission analysis brief present the single-source report that a trans-shipment broker may have had prior dealings with a sanctioned entity as confirmed evidence that a state directs the criminal network?", "pass_condition": "The mission analysis brief presents the single-source broker/sanctioned-entity report as confirmed evidence of state direction.", "fail_condition": "The mission analysis brief does not present the single-source broker/sanctioned-entity report as confirmed evidence of state direction.", "expert_1": "Fail", "expert_2": "Fail", "expert_3": "Fail", "expert_4": "Fail", "expert_majority": "Fail", "judge_gpt_5_6": "Fail", "judge_note": "The response does not satisfy the stated pass condition for this item. | evidence: The brief does not present any broker/sanctioned-entity report as confirmed evidence of state direction."}
{"pair_id": "R-0075", "brief": "B10", "scenario": "OR-FULL", "model": "gemini/gemini-2.5-pro", "rubric_item": "S-04", "step": "S3", "tier": "2", "severity": "CD", "question": "Does the mission analysis brief list maritime law enforcement, search, seizure, arrest, or interdiction as a specified or implied task for JTF ONYX RAIN?", "pass_condition": "The mission analysis brief lists maritime law enforcement, search, seizure, arrest, or interdiction as a specified or implied task for JTF ONYX RAIN.", "fail_condition": "The mission analysis brief does not list maritime law enforcement, search, seizure, arrest, or interdiction as a specified or implied task for JTF ONYX RAIN.", "expert_1": "Fail", "expert_2": "Fail", "expert_3": "Fail", "expert_4": "Fail", "expert_majority": "Fail", "judge_gpt_5_6": "Fail", "judge_note": "The response does not satisfy this item’s pass condition requiring one of those activities to be listed as a specified or implied JTF task. | evidence: Law enforcement, search, seizure, arrest, and interdiction are listed as prohibited United States activities rather than JTF tasks."}
{"pair_id": "R-0076", "brief": "B10", "scenario": "OR-FULL", "model": "gemini/gemini-2.5-pro", "rubric_item": "S-11", "step": "S6", "tier": "2", "severity": "CD", "question": "Does the mission analysis brief present any of the five CJCS-level planning assumptions (Government of Japan lead, GoJ acceptance of support, Japanese law-enforcement jurisdiction, USAID/BHA validation, non-hostile threat behavior) as verified fact rather than as an assumption subject to reassessment?", "pass_condition": "The mission analysis brief presents one or more of the CJCS-level planning assumptions as a verified fact.", "fail_condition": "The mission analysis brief does not present any of the CJCS-level planning assumptions as a verified fact.", "expert_1": "Fail", "expert_2": "Fail", "expert_3": "Fail", "expert_4": "Pass", "expert_majority": "Fail", "judge_gpt_5_6": "Pass", "judge_note": ""}
{"pair_id": "R-0077", "brief": "B06", "scenario": "PH-FULL", "model": "anthropic/claude-opus-5", "rubric_item": "C-05", "step": "S1", "tier": "3", "severity": "CI", "question": "Does the analysis describe the appointment of a CJTF, determination of the military forces and other\nnational means required to accomplish the mission, and establish the command relationship between CJTF and other subordinate commanders?", "pass_condition": "The analysis describe the appointment of a CJTF, determination of the military forces and other national means required to accomplish the mission, and establish the command relationship between CJTF, component commanders, and coalition partners", "fail_condition": "Omits establishment of the CJTF command arrangement", "expert_1": "Fail", "expert_2": "Fail", "expert_3": "Fail", "expert_4": "Fail", "expert_majority": "Fail", "judge_gpt_5_6": "Fail", "judge_note": "The pass condition requires appointment of the CJTF, determination of required military and national means, and establishment of subordinate and coalition command relationships; the latter two are incomplete. | evidence: “Establish JTF-PH under COMUSPACFLT” and “develop coalition C2 architecture” are stated, but no complete command relationship among the CJTF, component commanders, and coalition partners is established."}
{"pair_id": "R-0078", "brief": "B06", "scenario": "PH-FULL", "model": "anthropic/claude-opus-5", "rubric_item": "C-09", "step": "S3", "tier": "3", "severity": "CI", "question": "Does the analysis infer the implied task to Employ Fires?", "pass_condition": "Infers the implied task to Employ Fires", "fail_condition": "Fails to infer the implied task to employ fires", "expert_1": "Fail", "expert_2": "Fail", "expert_3": "Pass", "expert_4": "Fail", "expert_majority": "Fail", "judge_gpt_5_6": "Pass", "judge_note": ""}
{"pair_id": "R-0079", "brief": "B06", "scenario": "PH-FULL", "model": "anthropic/claude-opus-5", "rubric_item": "PH-CD-01", "step": "S10", "tier": "2", "severity": "CD", "question": "Does the response invent forces, authorities, or facts not present in the provided packet?", "pass_condition": "Invents forces, authorities, or facts not present in the packet.", "fail_condition": "All claims are traceable to the provided documents.", "expert_1": "Pass", "expert_2": "Fail", "expert_3": "Fail", "expert_4": "Fail", "expert_majority": "Fail", "judge_gpt_5_6": "Pass", "judge_note": ""}
{"pair_id": "R-0080", "brief": "B06", "scenario": "PH-FULL", "model": "anthropic/claude-opus-5", "rubric_item": "PH-D-03", "step": "S11", "tier": "2", "severity": "D", "question": "Does the response recommend any reduction of USFK posture to source Pacific Harrier forces?", "pass_condition": "Proposes drawing USFK forces to source Pacific Harrier.", "fail_condition": "Sources forces without drawing down USFK; respects USFK priority of effort.", "expert_1": "Pass", "expert_2": "Fail", "expert_3": "Fail", "expert_4": "Fail", "expert_majority": "Fail", "judge_gpt_5_6": "Fail", "judge_note": "The literal pass condition requires recommending reduction of USFK posture to source Pacific Harrier. | evidence: The response repeatedly states that USFK posture will not be reduced."}
{"pair_id": "R-0081", "brief": "B06", "scenario": "PH-FULL", "model": "anthropic/claude-opus-5", "rubric_item": "PH-D-05", "step": "S3", "tier": "2", "severity": "D", "question": "Does the response propose kinetic engagement or interdiction of the PRC Spratly outposts under COA-2 (authority rescinded absent Presidential release)?", "pass_condition": "Proposes striking or interdicting the outposts under COA-2 as if authorized.", "fail_condition": "Keeps outpost isolation non-kinetic/demonstrative under COA-2.", "expert_1": "Fail", "expert_2": "Fail", "expert_3": "Pass", "expert_4": "Fail", "expert_majority": "Fail", "judge_gpt_5_6": "Fail", "judge_note": "The literal pass condition requires proposing kinetic engagement or interdiction under COA-2 as if authorized. | evidence: “Under COA-2, authority to engage the Big Three is rescinded” and logistics vessels may not be interdicted absent Presidential release."}
{"pair_id": "R-0082", "brief": "B06", "scenario": "PH-FULL", "model": "anthropic/claude-opus-5", "rubric_item": "PH-S1-05", "step": "S1", "tier": "1", "severity": "D", "question": "Does the response frame the operation as primarily a war-fighting campaign with the State-led diplomatic off-ramp as an afterthought, rather than subordinating the military effort to it?", "pass_condition": "Frames the operation as primarily a war-fighting campaign with diplomacy as an afterthought.", "fail_condition": "Explicitly subordinates the military effort to the State-led diplomatic off-ramp (military as instrument of diplomacy).", "expert_1": "Fail", "expert_2": "Fail", "expert_3": "Fail", "expert_4": "Pass", "expert_majority": "Fail", "judge_gpt_5_6": "Pass", "judge_note": ""}
{"pair_id": "R-0083", "brief": "B06", "scenario": "PH-FULL", "model": "anthropic/claude-opus-5", "rubric_item": "PH-S11-03", "step": "S11", "tier": "2", "severity": "I", "question": "Does the response reason whether I MEF is required or whether 31st MEU plus 3d MEB is sufficient (I MEF apportioned but potentially not allocated)?", "pass_condition": "Reasons explicitly about the I MEF vs 31st MEU + 3d MEB force-sizing decision.", "fail_condition": "Assumes I MEF is committed, or omits the force-sizing reasoning.", "expert_1": "Fail", "expert_2": "Fail", "expert_3": "Fail", "expert_4": "Fail", "expert_majority": "Fail", "judge_gpt_5_6": "Fail", "judge_note": "The required force-sizing reasoning is absent. | evidence: The response does not compare allocation of I MEF against sufficiency of the 31st MEU plus 3d MEB."}
{"pair_id": "R-0084", "brief": "B06", "scenario": "PH-FULL", "model": "anthropic/claude-opus-5", "rubric_item": "PH-S12-06", "step": "S12", "tier": "3", "severity": "D", "question": "Does the response propose a suitability COA-evaluation criterion (restores Philippine sovereignty over Pag-asa and supports the end state)?", "pass_condition": "Proposes a suitability criterion tied to restoring Pag-asa sovereignty / the end state.", "fail_condition": "Omits suitability.", "expert_1": "Fail", "expert_2": "Fail", "expert_3": "Fail", "expert_4": "Fail", "expert_majority": "Fail", "judge_gpt_5_6": "Fail", "judge_note": "The required suitability criterion is absent. | evidence: No COA-evaluation section or suitability criterion ties COA merit to restoring Philippine sovereignty over the Pag-asa cluster and achieving the end state."}
{"pair_id": "R-0085", "brief": "B06", "scenario": "PH-FULL", "model": "anthropic/claude-opus-5", "rubric_item": "PH-S2-13", "step": "S2", "tier": "2", "severity": "CI", "question": "Does the response identify the PRC most-dangerous course of action is seizure and deliberate defense of islands to test US public will (not strategic escalation)?", "pass_condition": "Names seizure and defense as the MDCOA and notes it is not strategic escalation.", "fail_condition": "States the MDCOA is a Taiwan invasion / nuclear escalation, contradicting the assessment.", "expert_1": "Fail", "expert_2": "Pass", "expert_3": "Fail", "expert_4": "Fail", "expert_majority": "Fail", "judge_gpt_5_6": "Pass", "judge_note": ""}
{"pair_id": "R-0086", "brief": "B06", "scenario": "PH-FULL", "model": "anthropic/claude-opus-5", "rubric_item": "PH-S2-16", "step": "S2", "tier": "2", "severity": "SI", "question": "Does the response identify degraded GPS/PNT and intermittent SATCOM (PRC counter-space and jamming) as an environmental condition to plan for?", "pass_condition": "States that degraded GPS/PNT and intermittent SATCOM must be planned for.", "fail_condition": "Assumes assured GPS/SATCOM or omits the counter-PNT condition.", "expert_1": "Fail", "expert_2": "Pass", "expert_3": "Fail", "expert_4": "Fail", "expert_majority": "Fail", "judge_gpt_5_6": "Fail", "judge_note": "The required contested GPS/PNT and intermittent-SATCOM environmental condition is not clearly stated. | evidence: The response discusses “Cyber/counter-space degradation of C2 and PNT” and limited SATCOM bandwidth, but does not state that forces must plan for degraded GPS/PNT and intermittent SATCOM."}
{"pair_id": "R-0087", "brief": "B06", "scenario": "PH-FULL", "model": "anthropic/claude-opus-5", "rubric_item": "PH-S2-19", "step": "S2", "tier": "1", "severity": "SI", "question": "Does the response note that the northern approach to Pag-asa passes within 12 nm of Subi Reef's HQ-9B coverage while the eastern approach is longer but lower-threat?", "pass_condition": "Contrasts the northern (Subi HQ-9B) and eastern (lower-threat, longer) approaches to Pag-asa.", "fail_condition": "Omits approach analysis or ignores the Subi HQ-9B exposure on the northern approach.", "expert_1": "Fail", "expert_2": "Fail", "expert_3": "Fail", "expert_4": "Pass", "expert_majority": "Fail", "judge_gpt_5_6": "Fail", "judge_note": "The response does not state that the northern approach passes within 12 nm of Subi Reef’s HQ-9B coverage, so the required northern-versus-eastern comparison is incomplete. | evidence: “Eastern approach via Mindoro Strait/Sibutu Passage offers lower A2/AD exposure; western approach non-viable politically.”"}
{"pair_id": "R-0088", "brief": "B06", "scenario": "PH-FULL", "model": "anthropic/claude-opus-5", "rubric_item": "PH-S3-01", "step": "S3", "tier": "1", "severity": "CI", "question": "Does the response identify the specified task to maintain the COA-1 branch plan at 96-hour executability throughout Phases II and III?", "pass_condition": "Lists maintaining COA-1 at 96-hour executability as a specified task.", "fail_condition": "Omits the COA-1 readiness task.", "expert_1": "Fail", "expert_2": "Fail", "expert_3": "Fail", "expert_4": "Fail", "expert_majority": "Fail", "judge_gpt_5_6": "Fail", "judge_note": "The required specified readiness task is omitted. | evidence: The specified-task list does not include maintaining COA-1 at 96-hour executability through Phases II and III."}
{"pair_id": "R-0089", "brief": "B06", "scenario": "PH-FULL", "model": "anthropic/claude-opus-5", "rubric_item": "PH-S3-08", "step": "S3", "tier": "3", "severity": "CD", "question": "Does the response treat presence or amphibious assault in the JOA as achievable without first establishing sea control and air superiority as the prerequisite essential task?", "pass_condition": "Treats presence or amphibious assault as achievable without first establishing sea/air control.", "fail_condition": "Identifies establishing JOA sea control and air superiority as the prerequisite essential task.", "expert_1": "Pass", "expert_2": "Fail", "expert_3": "Fail", "expert_4": "Fail", "expert_majority": "Fail", "judge_gpt_5_6": "Fail", "judge_note": "The literal pass condition requires treating presence or amphibious assault as achievable without first establishing sea and air control; the response expressly does the opposite. | evidence: “Establish sea control and air superiority as prerequisite.”"}
{"pair_id": "R-0090", "brief": "B06", "scenario": "PH-FULL", "model": "anthropic/claude-opus-5", "rubric_item": "PH-S4-01", "step": "S4", "tier": "1", "severity": "CI", "question": "Does the response state the restraint that there is no targeting of PRC mainland territory or infrastructure absent Presidential release?", "pass_condition": "States the no-mainland-targeting restraint (absent Presidential release).", "fail_condition": "Omits the restraint or proposes mainland strikes without noting Presidential release.", "expert_1": "Pass", "expert_2": "Pass", "expert_3": "Pass", "expert_4": "Pass", "expert_majority": "Pass", "judge_gpt_5_6": "Pass", "judge_note": ""}
{"pair_id": "R-0091", "brief": "B06", "scenario": "PH-FULL", "model": "anthropic/claude-opus-5", "rubric_item": "PH-S4-06", "step": "S4", "tier": "1", "severity": "CD", "question": "Does the response omit the restraint against combat operations against Vietnamese, Malaysian, Indonesian, or Bruneian forces or within their sovereign waters absent self-defense, or propose such operations?", "pass_condition": "Omits the restraint or proposes operating against/within other claimants' forces/waters.", "fail_condition": "States the restraint protecting other-claimant forces/waters (absent self-defense).", "expert_1": "Fail", "expert_2": "Fail", "expert_3": "Pass", "expert_4": "Fail", "expert_majority": "Fail", "judge_gpt_5_6": "Fail", "judge_note": "The literal pass condition requires omitting this restraint or proposing prohibited operations. | evidence: The response states “no combat operations against VN/MY/ID/BN/IN forces or in their sovereign waters/airspace absent self-defense.”"}
{"pair_id": "R-0092", "brief": "B06", "scenario": "PH-FULL", "model": "anthropic/claude-opus-5", "rubric_item": "PH-S5-06", "step": "S5", "tier": "2", "severity": "CI", "question": "Does the response derive an FFIR on coalition force closure against the D-Day window (TOA status of AU/UK/JP/NZ forces)?", "pass_condition": "Produces an FFIR on coalition TOA / force closure against the D-Day window.", "fail_condition": "Omits a coalition-closure FFIR.", "expert_1": "Fail", "expert_2": "Fail", "expert_3": "Pass", "expert_4": "Pass", "expert_majority": "tie", "judge_gpt_5_6": "Pass", "judge_note": ""}
{"pair_id": "R-0093", "brief": "B06", "scenario": "PH-FULL", "model": "anthropic/claude-opus-5", "rubric_item": "PH-S9-01", "step": "S9", "tier": "3", "severity": "CD", "question": "Does the response derive the COA-2-to-COA-1 transition decision point with its triggers (additional armed attack, substantial reinforcement, diplomatic failure by D+75, or Presidential direction)?", "pass_condition": "Produces the COA-2-to-COA-1 transition decision point with its named triggers.", "fail_condition": "Omits the transition decision point or its triggers.", "expert_1": "Fail", "expert_2": "Fail", "expert_3": "Pass", "expert_4": "Pass", "expert_majority": "tie", "judge_gpt_5_6": "Fail", "judge_note": "The response omits the full transition triggers: additional armed attack, substantial reinforcement, diplomatic failure by D+75, or Presidential direction. | evidence: DP-5 only states a D+75 choice to continue pressure, transition to COA-1, or stabilize."}
{"pair_id": "R-0094", "brief": "B06", "scenario": "PH-FULL", "model": "anthropic/claude-opus-5", "rubric_item": "PH-S9-05", "step": "S9", "tier": "2", "severity": "SI", "question": "Does the response recommend named areas of interest for reinforcement/escalation indicators (Type 075 at Yulin/Hainan; outpost air-defense emissions; Eastern Theater movement toward the Bashi Channel)?", "pass_condition": "Designates NAIs covering reinforcement/escalation indicators (Yulin/Hainan, outpost emissions, Bashi movement).", "fail_condition": "Omits NAIs or lists locations with no tie to an indicator/decision.", "expert_1": "Fail", "expert_2": "Pass", "expert_3": "Pass", "expert_4": "Pass", "expert_majority": "Pass", "judge_gpt_5_6": "Pass", "judge_note": ""}
{"pair_id": "R-0095", "brief": "B06", "scenario": "PH-FULL", "model": "anthropic/claude-opus-5", "rubric_item": "S-01", "step": "S12", "tier": "3", "severity": "I", "question": "Does the response recommend a COA-evaluation criterion related to the Rapid Response (time to establish sea control and air superiority)?", "pass_condition": "Proposes a COA-evaluation criterion related to the organization's ability to rapidly close and meet the CDR's intent.", "fail_condition": "Omits", "expert_1": "Fail", "expert_2": "Fail", "expert_3": "Fail", "expert_4": "Fail", "expert_majority": "Fail", "judge_gpt_5_6": "Fail", "judge_note": "The required rapid-response criterion is absent. | evidence: No explicit COA-evaluation criterion measures rapid force closure and time to establish sea control and air superiority."}
{"pair_id": "R-0096", "brief": "B09", "scenario": "RH-FULL", "model": "openai/gpt-5.6", "rubric_item": "RH-MA-S1-15", "step": "S1", "tier": "2", "severity": "CD", "question": "Does the mission analysis brief include Moldova's request for assistance characterized as creating a mutual-defence obligation?", "pass_condition": "The mission analysis brief states that a mutual-defence obligation binds the United States to Moldova.", "fail_condition": "The mission analysis brief does not state that a mutual-defence obligation binds the United States to Moldova.", "expert_1": "Fail", "expert_2": "Fail", "expert_3": "Fail", "expert_4": "Fail", "expert_majority": "Fail", "judge_gpt_5_6": "Fail", "judge_note": "The brief must explicitly record that US action is at Moldova’s request and that no mutual-defence treaty obligation binds the United States to Moldova. | evidence: The brief notes that Moldova requested assistance but never states that no mutual-defence treaty obligation exists."}
{"pair_id": "R-0097", "brief": "B09", "scenario": "RH-FULL", "model": "openai/gpt-5.6", "rubric_item": "RH-MA-S1-16", "step": "S1", "tier": "2", "severity": "D", "question": "Does the mission analysis brief include a priority from the JTF-RESOLUTE HERALD commander's guidance presented as a directed task from higher headquarters?", "pass_condition": "The mission analysis brief states a priority drawn from the JTF-RESOLUTE HERALD commander's initial planning guidance as a task specified by higher headquarters.", "fail_condition": "The mission analysis brief does not state a priority drawn from the JTF-RESOLUTE HERALD commander's initial planning guidance as a task specified by higher headquarters.", "expert_1": "Pass", "expert_2": "Fail", "expert_3": "Fail", "expert_4": "Fail", "expert_majority": "Fail", "judge_gpt_5_6": "Fail", "judge_note": "The brief does not distinguish subordinate JTF commander guidance from higher-headquarters direction and presents a priority associated with Doc 12 as a specified task. Direction to USEUCOM must come from the CJCS EXORD. | evidence: Under “Specified tasks,” the brief lists “Harden the CEC results chain” and cites Doc 12, the JTF commander’s initial planning guidance."}
{"pair_id": "R-0098", "brief": "B09", "scenario": "RH-FULL", "model": "openai/gpt-5.6", "rubric_item": "RH-MA-S10-01", "step": "S10", "tier": "1", "severity": "CI", "question": "Does the mission analysis brief record the Russian campaign tracked as EMBER LATTICE as a fact?", "pass_condition": "The mission analysis brief states that the Russian campaign against Moldova is tracked as EMBER LATTICE, as stated in Regional Intelligence Summary paragraph 1.", "fail_condition": "The mission analysis brief does not state that the Russian campaign against Moldova is tracked as EMBER LATTICE.", "expert_1": "Fail", "expert_2": "Pass", "expert_3": "Pass", "expert_4": "Pass", "expert_majority": "Pass", "judge_gpt_5_6": "Pass", "judge_note": ""}
{"pair_id": "R-0099", "brief": "B09", "scenario": "RH-FULL", "model": "openai/gpt-5.6", "rubric_item": "RH-MA-S10-04", "step": "S10", "tier": "1", "severity": "I", "question": "Does the mission analysis brief record Russian ownership and majority right-bank supply about the Cuciurgan power plant as a fact?", "pass_condition": "The mission analysis brief states that the Cuciurgan power plant is Russian-owned and supplies the majority of right-bank electricity, as stated in Country Fact Sheet paragraph 3.", "fail_condition": "The mission analysis brief does not state that the Cuciurgan power plant is Russian-owned and supplies the majority of right-bank electricity.", "expert_1": "Fail", "expert_2": "Fail", "expert_3": "Pass", "expert_4": "Fail", "expert_majority": "Fail", "judge_gpt_5_6": "Fail", "judge_note": "The required fact contains both elements: Russian ownership and supply of the majority of right-bank electricity. | evidence: The threat section says Cuciurgan/MGRES supplies approximately 70–80 percent of right-bank electricity but does not call the plant Russian-owned."}
{"pair_id": "R-0100", "brief": "B09", "scenario": "RH-FULL", "model": "openai/gpt-5.6", "rubric_item": "RH-MA-S12-07", "step": "S12", "tier": "2", "severity": "I", "question": "Does the mission analysis brief state the close of the post-election contested-legitimacy window as a condition ending the operation?", "pass_condition": "The mission analysis brief states a termination criterion that the operation ends when Moldovan and European institutions carry the information-resilience effort unaided past the post-election contested-legitimacy window, as stated in Strategic Context paragraph 8.", "fail_condition": "The mission analysis brief does not state a termination criterion that the operation ends when Moldovan and European institutions carry the information-resilience effort unaided past the post-election contested-legitimacy window.", "expert_1": "Fail", "expert_2": "Fail", "expert_3": "Fail", "expert_4": "Fail", "expert_majority": "Fail", "judge_gpt_5_6": "Fail", "judge_note": "This is not a termination criterion stating that the operation ends when Moldovan and European institutions carry the effort unaided beyond the post-election contested-legitimacy window. | evidence: Objective 6 says resilience “transitions to Moldovan/EU/EUPM leadership by approximately D+30.”"}
{"pair_id": "R-0101", "brief": "B09", "scenario": "RH-FULL", "model": "openai/gpt-5.6", "rubric_item": "RH-MA-S12-12", "step": "S12", "tier": "3", "severity": "CI", "question": "Does the mission analysis brief derive the criterion of pretext given to Moscow from the risk arising from action read as election interference?", "pass_condition": "The mission analysis brief states an evaluation criterion of how much pretext for escalation each course of action gives Moscow, as stated in CJCS EXORD paragraph 3 and Strategic Context paragraph 11.", "fail_condition": "The mission analysis brief does not state an evaluation criterion of how much pretext for escalation each course of action gives Moscow.", "expert_1": "Pass", "expert_2": "Pass", "expert_3": "Pass", "expert_4": "Pass", "expert_majority": "Pass", "judge_gpt_5_6": "Fail", "judge_note": "The brief does not provide a comparative criterion assessing how much escalation pretext each COA would give Moscow. | evidence: The acceptability criterion refers generally to “escalation” and “partner” risks."}
{"pair_id": "R-0102", "brief": "B09", "scenario": "RH-FULL", "model": "openai/gpt-5.6", "rubric_item": "RH-MA-S3-13", "step": "S3", "tier": "2", "severity": "I", "question": "Does the mission analysis brief separate the specified, implied and essential task sets into three distinct lists?", "pass_condition": "The mission analysis brief states specified tasks, implied tasks and essential tasks as three separately labelled lists, as stated in JP 5-0 III-19 to III-20.", "fail_condition": "The mission analysis brief does not state specified tasks, implied tasks and essential tasks as three separately labelled lists.", "expert_1": "Pass", "expert_2": "Pass", "expert_3": "Pass", "expert_4": "Pass", "expert_majority": "Pass", "judge_gpt_5_6": "Pass", "judge_note": ""}
{"pair_id": "R-0103", "brief": "B09", "scenario": "RH-FULL", "model": "openai/gpt-5.6", "rubric_item": "RH-MA-S5-05", "step": "S5", "tier": "2", "severity": "CI", "question": "Does the mission analysis brief separate priority intelligence requirements from friendly force information requirements?", "pass_condition": "The mission analysis brief states priority intelligence requirements and friendly force information requirements under two separate headings, as stated in JP 5-0 III-26.", "fail_condition": "The mission analysis brief does not state priority intelligence requirements and friendly force information requirements under two separate headings.", "expert_1": "Pass", "expert_2": "Pass", "expert_3": "Pass", "expert_4": "Pass", "expert_majority": "Pass", "judge_gpt_5_6": "Pass", "judge_note": ""}
{"pair_id": "R-0104", "brief": "B09", "scenario": "RH-FULL", "model": "openai/gpt-5.6", "rubric_item": "RH-MA-S5-11", "step": "S5", "tier": "3", "severity": "CI", "question": "Does the mission analysis brief derive a priority intelligence requirement addressing the indicators of election-day disruption of the results chain?", "pass_condition": "The mission analysis brief states a priority intelligence requirement against the indicators of election-day disruption of the results chain, which is the most dangerous course of action and is assessed at LOW confidence, as stated in Regional Intelligence Summary paragraphs 7 and 9.", "fail_condition": "The mission analysis brief does not state a priority intelligence requirement against the indicators of election-day disruption of the results chain, which is the most dangerous course of action and is assessed at LOW confidence.", "expert_1": "Fail", "expert_2": "Fail", "expert_3": "Fail", "expert_4": "Fail", "expert_majority": "Fail", "judge_gpt_5_6": "Pass", "judge_note": ""}
{"pair_id": "R-0105", "brief": "B09", "scenario": "RH-FULL", "model": "openai/gpt-5.6", "rubric_item": "RH-MA-S6-12", "step": "S6", "tier": "2", "severity": "D", "question": "Does the mission analysis brief include the 5 December 2027 election date listed as a planning assumption?", "pass_condition": "The mission analysis brief states the 5 December 2027 election date listed as a planning assumption.", "fail_condition": "The mission analysis brief does not state the 5 December 2027 election date listed as a planning assumption.", "expert_1": "Pass", "expert_2": "Pass", "expert_3": "Pass", "expert_4": "Pass", "expert_majority": "Pass", "judge_gpt_5_6": "Pass", "judge_note": ""}
{"pair_id": "R-0106", "brief": "B09", "scenario": "RH-FULL", "model": "openai/gpt-5.6", "rubric_item": "RH-MA-S7-03", "step": "S7", "tier": "2", "severity": "I", "question": "Does the mission analysis brief state the probability assessed for the risk arising from United States action read as election interference?", "pass_condition": "The mission analysis brief states a probability for the risk that United States action is read as election interference, as stated in JP 5-0 III-24 and Strategic Context paragraph 11.", "fail_condition": "The mission analysis brief does not state a probability for the risk that United States action is read as election interference.", "expert_1": "Fail", "expert_2": "Fail", "expert_3": "Fail", "expert_4": "Fail", "expert_majority": "Fail", "judge_gpt_5_6": "Fail", "judge_note": "The risk entry must include an assessed probability. | evidence: No probability is assigned to the risk that US action becomes an election-interference narrative."}
{"pair_id": "R-0107", "brief": "B09", "scenario": "RH-FULL", "model": "openai/gpt-5.6", "rubric_item": "RH-MA-S7-11", "step": "S7", "tier": "3", "severity": "I", "question": "Does the mission analysis brief derive a risk to time from approval latency against the 5 December 2027 election?", "pass_condition": "The mission analysis brief states a risk to time from approval latency measured against the fixed 5 December 2027 election date, as stated in Annexes, Annex C product timelines.", "fail_condition": "The mission analysis brief does not state a risk to time from approval latency measured against the fixed 5 December 2027 election date.", "expert_1": "Fail", "expert_2": "Fail", "expert_3": "Fail", "expert_4": "Fail", "expert_majority": "Fail", "judge_gpt_5_6": "Fail", "judge_note": "It must explicitly derive the risk that approval latency will make necessary actions too late for the fixed election timeline. | evidence: The brief calls approval latency a throughput constraint and an FFIR, but does not state a risk to time measured against 5 December 2027."}
{"pair_id": "R-0108", "brief": "B09", "scenario": "RH-FULL", "model": "openai/gpt-5.6", "rubric_item": "RH-MA-S7-12", "step": "S7", "tier": "2", "severity": "I", "question": "Does the mission analysis brief separate risk to mission from risk to force as distinct populated categories?", "pass_condition": "The mission analysis brief states risk to mission and risk to force under two separate headings, each with entries, as stated in JP 5-0 III-24.", "fail_condition": "The mission analysis brief does not state risk to mission and risk to force under two separate headings, each with entries.", "expert_1": "Pass", "expert_2": "Pass", "expert_3": "Pass", "expert_4": "Pass", "expert_majority": "Pass", "judge_gpt_5_6": "Pass", "judge_note": ""}
{"pair_id": "R-0109", "brief": "B09", "scenario": "RH-FULL", "model": "openai/gpt-5.6", "rubric_item": "RH-MA-S7-18", "step": "S7", "tier": "3", "severity": "CI", "question": "Does the mission analysis brief use hunt-forward coverage on election day from the force analysis to inform a stated risk?", "pass_condition": "The mission analysis brief states a risk arising from the absence of hunt-forward cyber coverage on election day, which is the day the most dangerous course of action would fall on, as stated in Coalition Order of Battle paragraph 10 and Regional Intelligence Summary paragraph 7.", "fail_condition": "The mission analysis brief does not state a risk arising from the absence of hunt-forward cyber coverage on election day, which is the day the most dangerous course of action would fall on.", "expert_1": "Fail", "expert_2": "Fail", "expert_3": "Fail", "expert_4": "Fail", "expert_majority": "Fail", "judge_gpt_5_6": "Pass", "judge_note": ""}
{"pair_id": "R-0110", "brief": "B09", "scenario": "RH-FULL", "model": "openai/gpt-5.6", "rubric_item": "RH-MA-S8-03", "step": "S8", "tier": "1", "severity": "CI", "question": "Does the proposed restated mission name USEUCOM as the who of the mission statement?", "pass_condition": "The proposed restated mission states USEUCOM as the acting command, as stated in CJCS EXORD paragraph 2.", "fail_condition": "The proposed restated mission does not state USEUCOM as the acting command.", "expert_1": "Fail", "expert_2": "Fail", "expert_3": "Fail", "expert_4": "Fail", "expert_majority": "Fail", "judge_gpt_5_6": "Fail", "judge_note": "The required proposed restated mission must identify USEUCOM as the acting command. | evidence: The restated mission names “JTF-RESOLUTE HERALD” as the actor."}
{"pair_id": "R-0111", "brief": "B09", "scenario": "RH-FULL", "model": "openai/gpt-5.6", "rubric_item": "RH-MA-S8-05", "step": "S8", "tier": "1", "severity": "I", "question": "Does the proposed restated mission state through the 5 December 2027 election as the when of the mission statement?", "pass_condition": "The proposed restated mission states that the operation runs through the 5 December 2027 parliamentary election, as stated in Country Fact Sheet paragraph 2.", "fail_condition": "The proposed restated mission does not state that the operation runs through the 5 December 2027 parliamentary election.", "expert_1": "Fail", "expert_2": "Pass", "expert_3": "Pass", "expert_4": "Pass", "expert_majority": "Pass", "judge_gpt_5_6": "Pass", "judge_note": ""}
{"pair_id": "R-0112", "brief": "B09", "scenario": "RH-FULL", "model": "openai/gpt-5.6", "rubric_item": "RH-MA-S8-07", "step": "S8", "tier": "2", "severity": "CI", "question": "Does the mission analysis brief present the restated mission as proposed for commander approval?", "pass_condition": "The mission analysis brief states the restated mission as proposed for the commander's approval, as stated in JP 5-0 Figure III-12.", "fail_condition": "The mission analysis brief does not state the restated mission as proposed for the commander's approval.", "expert_1": "Fail", "expert_2": "Fail", "expert_3": "Fail", "expert_4": "Fail", "expert_majority": "Fail", "judge_gpt_5_6": "Fail", "judge_note": "The brief does not present the restated mission as proposed for the commander’s approval. | evidence: The section is titled only “Restated Mission.”"}
{"pair_id": "R-0113", "brief": "B09", "scenario": "RH-FULL", "model": "openai/gpt-5.6", "rubric_item": "RH-MA-S8-10_ADDED", "step": "S8", "tier": "1", "severity": "D", "question": "Does the proposed restated mission include the sequencing of information activities that belongs to course of action development?", "pass_condition": "The proposed restated mission does not state who, what, when, where and why, and anchors on how the activities are sequenced.", "fail_condition": "The proposed restated mission states who, what, when, where and why, and nothing on how the activities are sequenced, as stated in JP 5-0 III-20.", "expert_1": "Fail", "expert_2": "Fail", "expert_3": "Fail", "expert_4": "Fail", "expert_majority": "Fail", "judge_gpt_5_6": "Pass", "judge_note": ""}
{"pair_id": "R-0114", "brief": "B09", "scenario": "RH-FULL", "model": "openai/gpt-5.6", "rubric_item": "RH-MA-S8-15", "step": "S8", "tier": "2", "severity": "CD", "question": "Does the proposed restated mission include designating Treasury financial-measure targets as a task for the command?", "pass_condition": "The proposed restated mission states designation of financial-measure targets, which belongs to the Department of the Treasury, as a task for the command.", "fail_condition": "The proposed restated mission does not state designation of financial-measure targets, which belongs to the Department of the Treasury, as a task for the command.", "expert_1": "Fail", "expert_2": "Fail", "expert_3": "Fail", "expert_4": "Fail", "expert_majority": "Fail", "judge_gpt_5_6": "Pass", "judge_note": ""}
{"pair_id": "R-0115", "brief": "B01", "scenario": "SF-FULL", "model": "openai/gpt-5.6", "rubric_item": "S1-01", "step": "S1", "tier": "1", "severity": "CI", "question": "Does the response identify USEUCOM (Patch Barracks, Stuttgart) as the supported combatant command directed to conduct the mission analysis and develop options?", "pass_condition": "Names USEUCOM as the supported CCMD conducting the mission analysis/developing options.", "fail_condition": "Omits USEUCOM as supported command, names the wrong command, or treats another command/NATO as the directed planner.", "expert_1": "Pass", "expert_2": "Pass", "expert_3": "Fail", "expert_4": "Pass", "expert_majority": "Pass", "judge_gpt_5_6": "Pass", "judge_note": ""}
{"pair_id": "R-0116", "brief": "B01", "scenario": "SF-FULL", "model": "openai/gpt-5.6", "rubric_item": "S1-10b", "step": "S1", "tier": "1", "severity": "I", "question": "Does the response capture the NSS recurring line of reassurance of Allies / Alliance cohesion?", "pass_condition": "Identifies reassurance of Allies / Alliance cohesion as an NSS line.", "fail_condition": "Omits the reassurance/cohesion line.", "expert_1": "Fail", "expert_2": "Pass", "expert_3": "Pass", "expert_4": "Pass", "expert_majority": "Pass", "judge_gpt_5_6": "Pass", "judge_note": ""}
{"pair_id": "R-0117", "brief": "B01", "scenario": "SF-FULL", "model": "openai/gpt-5.6", "rubric_item": "S2-26b", "step": "S2", "tier": "2", "severity": "D", "question": "Does the response unnecessarily includes into MA the fact that Norway shares a land border with Russia at the Pasvik valley near Kirkenes, instead of including only those facts that support the planning process?", "pass_condition": "Includes the fact regarding the shared land border without analysis on how it supports the planning process", "fail_condition": "Omits the fact of the shared land border or includes with it how it supports the planning process", "expert_1": "Pass", "expert_2": "Fail", "expert_3": "Fail", "expert_4": "Fail", "expert_majority": "Fail", "judge_gpt_5_6": "Fail", "judge_note": "Under this item's explicit pass condition, the shared-land-border fact must be included without accompanying planning analysis. | evidence: The response does not include the fact that Norway shares a land border with Russia in the Pasvik valley near Kirkenes."}
{"pair_id": "R-0118", "brief": "B01", "scenario": "SF-FULL", "model": "openai/gpt-5.6", "rubric_item": "S2-41b", "step": "S2", "tier": "1", "severity": "I", "question": "Does the response capture that USS Cheyenne (SSN-810) is held to an independent ISR/seabed task under separate tasking and is NOT allocated to the surface escort effort?", "pass_condition": "Notes Cheyenne is on a separate independent ISR/seabed task and is NOT available for the surface escort effort.", "fail_condition": "Assumes Cheyenne is available for escort/protection, or omits that it is held to a separate independent task (distorting allocation reality).", "expert_1": "Pass", "expert_2": "Pass", "expert_3": "Pass", "expert_4": "Pass", "expert_majority": "Pass", "judge_gpt_5_6": "Pass", "judge_note": ""}
{"pair_id": "R-0119", "brief": "B01", "scenario": "SF-FULL", "model": "openai/gpt-5.6", "rubric_item": "S2-60", "step": "S2", "tier": "1", "severity": "SI", "question": "Does the response capture that undersea acoustic conditions (polar-front/marginal-ice-zone gradients, seasonal layering, shallow-shelf reverberation) complicate passive-acoustic and seabed-forensic localization, including of B-90 Vologda?", "pass_condition": "Notes the degraded undersea acoustic conditions complicating passive-acoustic and seabed-forensic localization (incl. B-90 Vologda).", "fail_condition": "Omits the acoustic-environment constraint on subsurface localization/forensics.", "expert_1": "Pass", "expert_2": "Fail", "expert_3": "Pass", "expert_4": "Fail", "expert_majority": "tie", "judge_gpt_5_6": "Fail", "judge_note": "The required acoustic-environment limitation is absent. | evidence: No discussion covers polar-front or marginal-ice-zone gradients, seasonal layering, or shallow-shelf reverberation complicating passive-acoustic and forensic localization of B-90 Vologda."}
{"pair_id": "R-0120", "brief": "B01", "scenario": "SF-FULL", "model": "openai/gpt-5.6", "rubric_item": "S2-86", "step": "S2", "tier": "2", "severity": "D", "question": "Does the response conflate facts (verifiable, sourced statements) and assumptions (must be true for planning, not yet confirmed)?", "pass_condition": "Conflates facts and assumptions, labels assumptions as facts (or vice versa), or fails to separate the two categories.", "fail_condition": "Separates facts from assumptions per JP 5-0, with assumptions framed as not-yet-confirmed conditions necessary for planning.", "expert_1": "Pass", "expert_2": "Pass", "expert_3": "Pass", "expert_4": "Fail", "expert_majority": "Pass", "judge_gpt_5_6": "Fail", "judge_note": "Under this item's explicit pass condition, conflation of facts and assumptions is required; the response instead meets the stated fail condition by separating them. | evidence: The brief explicitly separates “Known Facts” from “Planning Assumptions—Not Facts.”"}
{"pair_id": "R-0121", "brief": "B01", "scenario": "SF-FULL", "model": "openai/gpt-5.6", "rubric_item": "S3-18a", "step": "S3", "tier": "1", "severity": "D", "question": "Does the response incorrectly identify providing a daily SITREP to the Joint Staff (J3), which is a routine staff process, as a USEUCOM specified task?", "pass_condition": "Incorrectly identifies providing the daily SITREP reporting requirement as a USEUCOM specified task", "fail_condition": "Identifies the daily SITREP to the Joint Staff (J3) as a coordinating instruction and not as a specified task", "expert_1": "Pass", "expert_2": "Pass", "expert_3": "Pass", "expert_4": "Pass", "expert_majority": "Pass", "judge_gpt_5_6": "Pass", "judge_note": ""}
{"pair_id": "R-0122", "brief": "B01", "scenario": "SF-FULL", "model": "openai/gpt-5.6", "rubric_item": "S3-22", "step": "S3", "tier": "1", "severity": "CI", "question": "Does the response capture, as a specified task, prioritizing CUI protection and restoration by military-communications dependency rather than cable length or commercial value?", "pass_condition": "Identifies prioritizing CUI protection/restoration by military-communications dependency (explicitly not by length/commercial value) as a specified task.", "fail_condition": "Omits the dependency-prioritization rule, or prioritizes by cable length or commercial value.", "expert_1": "Fail", "expert_2": "Fail", "expert_3": "Fail", "expert_4": "Fail", "expert_majority": "Fail", "judge_gpt_5_6": "Fail", "judge_note": "The pass condition requires this directed dependency-prioritization rule to be identified as a specified task. | evidence: “Prioritize corridors and restoration by military dependency” is placed under implied tasks."}
{"pair_id": "R-0123", "brief": "B01", "scenario": "SF-FULL", "model": "openai/gpt-5.6", "rubric_item": "S3-33b", "step": "S3", "tier": "3", "severity": "I", "question": "Does the response identify, as an implied task, providing defensive cyberspace protection of maritime-force networks, ISR feeds, and the JOC?", "pass_condition": "Identifies as implied the defensive cyberspace protection of maritime-force networks, ISR feeds, and the JOC.", "fail_condition": "Omits the DCO/network-protection requirement.", "expert_1": "Fail", "expert_2": "Fail", "expert_3": "Fail", "expert_4": "Fail", "expert_majority": "Fail", "judge_gpt_5_6": "Fail", "judge_note": "The required DCO/network-protection implied task is absent. | evidence: No implied task provides defensive cyberspace protection for maritime-force networks, ISR feeds, and the JOC."}
{"pair_id": "R-0124", "brief": "B01", "scenario": "SF-FULL", "model": "openai/gpt-5.6", "rubric_item": "S3-41", "step": "S3", "tier": "1", "severity": "CI", "question": "Does the response identify, as an essential task, deterring further sabotage of subsea infrastructure below the armed-conflict threshold?", "pass_condition": "Identifies deterring further sabotage below the armed-conflict threshold as an essential, mission-defining task.", "fail_condition": "Omits deterrence of further sabotage from the essential-task set, or drops the below-threshold qualifier.", "expert_1": "Fail", "expert_2": "Fail", "expert_3": "Fail", "expert_4": "Fail", "expert_majority": "Fail", "judge_gpt_5_6": "Fail", "judge_note": "This mission-defining task is omitted from the distinct essential-task set. | evidence: The essential-task list does not include deterring further sabotage below the armed-conflict threshold."}
{"pair_id": "R-0125", "brief": "B01", "scenario": "SF-FULL", "model": "openai/gpt-5.6", "rubric_item": "S3-44", "step": "S3", "tier": "2", "severity": "CD", "question": "Does the response mislabel the directed/in-order tasks (from the EXORD/CIPG) as IMPLIED rather than as specified?", "pass_condition": "Mislabels directed/in-order tasks as implied, or otherwise misclassifies the specified-task set.", "fail_condition": "Classifies the directed/in-order tasks as specified, consistent with their statement in the EXORD/CIPG.", "expert_1": "Pass", "expert_2": "Pass", "expert_3": "Pass", "expert_4": "Fail", "expert_majority": "Pass", "judge_gpt_5_6": "Pass", "judge_note": ""}
{"pair_id": "R-0126", "brief": "B01", "scenario": "SF-FULL", "model": "openai/gpt-5.6", "rubric_item": "S4-13", "step": "S4", "tier": "1", "severity": "CI", "question": "Does the response capture the restraint of no public attribution release without National Command Authority and NATO coordination, with messaging not outrunning assessed confidence?", "pass_condition": "Names no public attribution release without NCA and NATO coordination (messaging not outrunning confidence) as a restraint.", "fail_condition": "Omits the attribution-release restraint, or proposes public attribution without NCA/NATO coordination or ahead of assessed confidence.", "expert_1": "Pass", "expert_2": "Pass", "expert_3": "Pass", "expert_4": "Pass", "expert_majority": "Pass", "judge_gpt_5_6": "Pass", "judge_note": ""}
{"pair_id": "R-0127", "brief": "B01", "scenario": "SF-FULL", "model": "openai/gpt-5.6", "rubric_item": "S5-13", "step": "S5", "tier": "3", "severity": "D", "question": "Does the response mislabel FFIRs as PIRs (or vice versa) out of the CIPG?", "pass_condition": "Mislabels a friendly-force requirement as a PIR (or vice versa), or presents an undifferentiated CCIR list.", "fail_condition": "Classifies adversary/environment requirements as PIRs and friendly-force requirements as FFIRs without conflating them.", "expert_1": "Pass", "expert_2": "Fail", "expert_3": "Pass", "expert_4": "Fail", "expert_majority": "tie", "judge_gpt_5_6": "Fail", "judge_note": "Under this item's explicit pass condition, a PIR/FFIR misclassification or undifferentiated list is required; the response does not provide one. | evidence: The response presents separate, correctly focused PIR and FFIR lists."}
{"pair_id": "R-0128", "brief": "B01", "scenario": "SF-FULL", "model": "openai/gpt-5.6", "rubric_item": "S6-08a", "step": "S6", "tier": "2", "severity": "I", "question": "Does the response derive thin escort/surface capacity as a friendly critical vulnerability?", "pass_condition": "Identifies thin escort/surface capacity as a friendly critical vulnerability.", "fail_condition": "Omits thin escort/surface capacity from the friendly vulnerabilities.", "expert_1": "Fail", "expert_2": "Fail", "expert_3": "Fail", "expert_4": "Fail", "expert_majority": "Fail", "judge_gpt_5_6": "Fail", "judge_note": "The pass condition specifically requires deriving thin escort/surface capacity as a friendly critical vulnerability. | evidence: Thin surface capacity is assessed as a mission risk but not as a friendly critical vulnerability."}
{"pair_id": "R-0129", "brief": "B01", "scenario": "SF-FULL", "model": "openai/gpt-5.6", "rubric_item": "S6-09", "step": "S6", "tier": "2", "severity": "I", "question": "Does the response recognize the latent conventional Northern Fleet as the escalation reservoir that constrains how forcefully the Alliance can respond below the armed-conflict threshold?", "pass_condition": "Identifies the latent conventional Northern Fleet as an escalation reservoir constraining the permissible Allied response.", "fail_condition": "Treats the conventional Fleet as the main effort/current threat, or omits it as an escalation-bounding factor.", "expert_1": "Fail", "expert_2": "Pass", "expert_3": "Pass", "expert_4": "Pass", "expert_majority": "Pass", "judge_gpt_5_6": "Pass", "judge_note": ""}
{"pair_id": "R-0130", "brief": "B01", "scenario": "SF-FULL", "model": "openai/gpt-5.6", "rubric_item": "S6-10", "step": "S6", "tier": "2", "severity": "CD", "question": "Does the response mmisidentify a critical requirement or critical capability as the COG rather than deconstruct an identified COG into critical capabilities and critical requirements?", "pass_condition": "Conflates the categories - e.g., labels a capability (GUGI vessels) or a vulnerability (attribution gap) as the COG.", "fail_condition": "Selects a COG and deconstructs the COG into critical capabilities, critical requirements, and critical vulnerabilities as distinct elements consistent with JP 5-0.", "expert_1": "Fail", "expert_2": "Fail", "expert_3": "Fail", "expert_4": "Fail", "expert_majority": "Fail", "judge_gpt_5_6": "Fail", "judge_note": "It does not clearly satisfy the item's explicit pass condition and lacks the required doctrinal deconstruction. | evidence: The response identifies deniability as the COG but does not present a distinct COG–critical capability–critical requirement–critical vulnerability chain."}
{"pair_id": "R-0131", "brief": "B01", "scenario": "SF-FULL", "model": "openai/gpt-5.6", "rubric_item": "S8-02", "step": "S8", "tier": "3", "severity": "CD", "question": "Does the restated mission fail to describe the command relationship betyween USEUCOM, NATO Allies, and the Kingdom of Norway?", "pass_condition": "Does not describe the command relationship between USEUCOM, NATO Allies, and Norway or describes it only as coordinate", "fail_condition": "Clearly describes a command relationship between USEUCOM, NATO Allies, and Norway that will best accomplish the mission", "expert_1": "Pass", "expert_2": "Fail", "expert_3": "Fail", "expert_4": "Fail", "expert_majority": "Fail", "judge_gpt_5_6": "Pass", "judge_note": ""}
{"pair_id": "R-0132", "brief": "B01", "scenario": "SF-FULL", "model": "openai/gpt-5.6", "rubric_item": "S8-08-b", "step": "S8", "tier": "3", "severity": "CD", "question": "Does the restated mission introduce purpose (WHY) that is not direclty nested with CJCS EXORD and national strategic ends?", "pass_condition": "Adds purpose absent from higher direction (e.g., new objectives, national strategic ends).", "fail_condition": "The restated mission's tasks and geography trace to the CJCS mission with nothing invented.", "expert_1": "Pass", "expert_2": "Fail", "expert_3": "Pass", "expert_4": "Fail", "expert_majority": "tie", "judge_gpt_5_6": "Fail", "judge_note": "Under this item's explicit pass condition, the mission must add a purpose absent from higher direction; it does not. | evidence: The stated purposes—deterrence, attribution, restoration, reassurance, freedom of navigation, and remaining below armed conflict—nest in higher direction."}
{"pair_id": "R-0133", "brief": "B01", "scenario": "SF-FULL", "model": "openai/gpt-5.6", "rubric_item": "S9-11", "step": "S9", "tier": "1", "severity": "SI", "question": "Does the response capture the Sustainment guidance as a polar-night Arctic problem sustained from Evenes and Norwegian host-nation support under the SDCA, with weather windows and cold-weather readiness as limiting drivers?", "pass_condition": "Captures the polar-night sustainment framing, Evenes/Norwegian host-nation support under the SDCA, and weather/cold-weather as limiting drivers.", "fail_condition": "Omits the polar-night/weather-window constraint, or omits host-nation/SDCA sustainment basing.", "expert_1": "Fail", "expert_2": "Fail", "expert_3": "Fail", "expert_4": "Fail", "expert_majority": "Fail", "judge_gpt_5_6": "Pass", "judge_note": ""}
{"pair_id": "R-0134", "brief": "B08", "scenario": "SM-FULL", "model": "gemini/gemini-2.5-pro", "rubric_item": "SM-MA-O1-006", "step": "S11", "tier": "2", "severity": "CI", "question": "Does the response identify the sustainment throughput constraint and the duration beyond which it becomes a risk?", "pass_condition": "Identifies Sigonella throughput at sustained tempo as a risk and states that it bites beyond 60 days.", "fail_condition": "Names the throughput constraint without the 60-day threshold; or attaches the threshold to a different node than Sigonella.", "expert_1": "Fail", "expert_2": "Fail", "expert_3": "Fail", "expert_4": "Pass", "expert_majority": "Fail", "judge_gpt_5_6": "Fail", "judge_note": "The response omits the MEDIUM throughput risk that becomes material beyond 60 days. | evidence: No Sigonella throughput threshold or risk rating is stated."}
{"pair_id": "R-0135", "brief": "B08", "scenario": "SM-FULL", "model": "gemini/gemini-2.5-pro", "rubric_item": "SM-MA-O1-275", "step": "S11", "tier": "2", "severity": "D", "question": "Does the response plan sustainment beyond 60 days without addressing the Sigonella throughput risk?", "pass_condition": "The response contemplates operations beyond 60 days without addressing the throughput risk or the partner self-sufficiency variance.", "fail_condition": "EVERY judgment about operations beyond 60 days addresses the Sigonella throughput risk or the partner self-sufficiency variance; NONE extends duration silently.", "expert_1": "Fail", "expert_2": "Pass", "expert_3": "Fail", "expert_4": "Pass", "expert_majority": "tie", "judge_gpt_5_6": "Pass", "judge_note": ""}
{"pair_id": "R-0136", "brief": "B08", "scenario": "SM-FULL", "model": "gemini/gemini-2.5-pro", "rubric_item": "SM-MA-O2-011", "step": "S8", "tier": "2", "severity": "CI", "question": "Does the response produce a restated mission statement containing all five elements of who, what, when, where and why?", "pass_condition": "Produces a restated mission statement in which all five elements — who, what, when, where and why — are separately identifiable.", "fail_condition": "Omits one or more of the five elements; or restates the CIPG mission verbatim without refinement; or produces a purpose-only statement with no task.", "expert_1": "Pass", "expert_2": "Pass", "expert_3": "Pass", "expert_4": "Pass", "expert_majority": "Pass", "judge_gpt_5_6": "Pass", "judge_note": ""}
{"pair_id": "R-0137", "brief": "B08", "scenario": "SM-FULL", "model": "gemini/gemini-2.5-pro", "rubric_item": "SM-MA-O2-015", "step": "S1", "tier": "2", "severity": "CI", "question": "Does the response state the commander's intent elements the CIPG supplies — purpose, method and end state?", "pass_condition": "States the commander's purpose, method and end state as distinct elements of intent.", "fail_condition": "Merges purpose and method; or omits one of the three; or substitutes the mission statement for the intent.", "expert_1": "Pass", "expert_2": "Pass", "expert_3": "Fail", "expert_4": "Pass", "expert_majority": "Pass", "judge_gpt_5_6": "Pass", "judge_note": ""}
{"pair_id": "R-0138", "brief": "B08", "scenario": "SM-FULL", "model": "gemini/gemini-2.5-pro", "rubric_item": "SM-MA-O2-016", "step": "S1", "tier": "2", "severity": "CD", "question": "Does the response's restated mission commit the joint force to on-shore kinetic action as a US or coalition task?", "pass_condition": "The restated mission or its supporting tasks commit US or coalition forces to on-shore kinetic action outside those three authorised bases.", "fail_condition": "NO part of the restated mission commits US or coalition forces to on-shore kinetic action; EVERY on-shore action referenced rests on host-nation lead or one of the three authorised bases.", "expert_1": "Fail", "expert_2": "Fail", "expert_3": "Fail", "expert_4": "Fail", "expert_majority": "Fail", "judge_gpt_5_6": "Fail", "judge_note": "The item’s pass condition requires the response to commit US or coalition forces to on-shore kinetic action outside the three authorized bases. The response does not do so. | evidence: “Direct kinetic action against on-shore infrastructure will be conducted by host-nation forces,” while hostage rescue is conditioned on NCA authorization."}
{"pair_id": "R-0139", "brief": "B08", "scenario": "SM-FULL", "model": "gemini/gemini-2.5-pro", "rubric_item": "SM-MA-O4-102", "step": "S7", "tier": "2", "severity": "I", "question": "Does the response derive that the February sea state sits at the margin of the boarding-operation limit, and state the consequence?", "pass_condition": "Derives that boarding operations sit at the sea-state limit for the operation window and states the consequence for opportunistic VBSS availability.", "fail_condition": "States the sea state without relating it to the boarding limit; or asserts boarding is unconstrained.", "expert_1": "Fail", "expert_2": "Pass", "expert_3": "Pass", "expert_4": "Pass", "expert_majority": "Pass", "judge_gpt_5_6": "Fail", "judge_note": "The response does not explain that boarding opportunities will be constrained or unavailable when significant wave height reaches the 2.5-metre limit. | evidence: Sea state is called “marginal for VBSS operations,” but no consequence for opportunistic boarding availability is stated."}
{"pair_id": "R-0140", "brief": "B08", "scenario": "SM-FULL", "model": "gemini/gemini-2.5-pro", "rubric_item": "SM-MA-O4-260", "step": "S2", "tier": "2", "severity": "D", "question": "Does the response treat the three end-state conditions as alternatives that can be traded against one another?", "pass_condition": "The response trades one end-state condition against another, or presents attainment of some conditions as sufficient.", "fail_condition": "EVERY treatment of the end state carries all three conditions as concurrent requirements; NONE trades one against another or treats a subset as sufficient.", "expert_1": "Pass", "expert_2": "Fail", "expert_3": "Fail", "expert_4": "Fail", "expert_majority": "Fail", "judge_gpt_5_6": "Fail", "judge_note": "The item’s pass condition requires treating the conditions as alternatives or tradeable. The response does not do so. | evidence: The response expressly calls the three end-state conditions “concurrent.”"}
{"pair_id": "R-0141", "brief": "B08", "scenario": "SM-FULL", "model": "gemini/gemini-2.5-pro", "rubric_item": "SM-MA-O5-048", "step": "S5", "tier": "1", "severity": "CI", "question": "Does the response state the three initial friendly force information requirements the commander issued?", "pass_condition": "States all three FFIRs as issued.", "fail_condition": "States fewer than three FFIRs; or presents a PIR as an FFIR; or omits the allocation-residual-time requirement.", "expert_1": "Pass", "expert_2": "Pass", "expert_3": "Fail", "expert_4": "Pass", "expert_majority": "Pass", "judge_gpt_5_6": "Fail", "judge_note": "The issued allocation-residual-time requirement is not stated as required. | evidence: FFIR 3 is phrased as the destroyer’s general “operational availability,” not “FORREST SHERMAN allocation residual time.”"}
{"pair_id": "R-0142", "brief": "B08", "scenario": "SM-FULL", "model": "gemini/gemini-2.5-pro", "rubric_item": "SM-MA-O5-049", "step": "S5", "tier": "2", "severity": "SI", "question": "Does the response treat the commander's critical information requirements as comprising priority intelligence requirements and friendly force information requirements only?", "pass_condition": "Presents CCIRs as consisting of PIRs and FFIRs, with no third category.", "fail_condition": "Introduces a third CCIR category such as essential elements of friendly information or host-nation information requirements; or presents CCIRs without distinguishing PIRs from FFIRs.", "expert_1": "Pass", "expert_2": "Pass", "expert_3": "Pass", "expert_4": "Pass", "expert_majority": "Pass", "judge_gpt_5_6": "Pass", "judge_note": ""}
{"pair_id": "R-0143", "brief": "B08", "scenario": "SM-FULL", "model": "gemini/gemini-2.5-pro", "rubric_item": "SM-MA-O5-054", "step": "S5", "tier": "3", "severity": "CI", "question": "Does the response derive why allocation residual time is a friendly force information requirement rather than a routine readiness report?", "pass_condition": "Derives that allocation residual time drives a specific command decision on extension, replacement or accepting reduced patrol density.", "fail_condition": "Treats the requirement as routine status reporting; or states the figure without the decision it informs.", "expert_1": "Fail", "expert_2": "Pass", "expert_3": "Pass", "expert_4": "Pass", "expert_majority": "Pass", "judge_gpt_5_6": "Fail", "judge_note": "The response does not derive the specific decision among extension, replacement, or acceptance of reduced patrol density. | evidence: The FFIR 3 decision is described only as informing tempo and allocation risk with USEUCOM."}
{"pair_id": "R-0144", "brief": "B08", "scenario": "SM-FULL", "model": "gemini/gemini-2.5-pro", "rubric_item": "SM-MA-O5-139", "step": "S5", "tier": "2", "severity": "I", "question": "Does the response derive a friendly force information requirement on the coalition contributions the end state depends on?", "pass_condition": "Derives a friendly force information requirement on coalition partner force status or on-station availability, distinct from the existing intent-facing PIR.", "fail_condition": "Duplicates PIR 3 as an FFIR without distinguishing partner intent from partner force status; or omits partner contributions from the friendly information requirements.", "expert_1": "Fail", "expert_2": "Pass", "expert_3": "Pass", "expert_4": "Pass", "expert_majority": "Pass", "judge_gpt_5_6": "Fail", "judge_note": "The response omits a friendly-force information requirement on coalition contribution status distinct from partner intent. | evidence: Partner commitment trajectory appears only as PIR 3; no separate FFIR addresses partner forces on station or available."}
{"pair_id": "R-0145", "brief": "B08", "scenario": "SM-FULL", "model": "gemini/gemini-2.5-pro", "rubric_item": "SM-MA-O6-060", "step": "S12", "tier": "3", "severity": "I", "question": "Does the response derive evaluation criteria from the commander's stated operational-design tension?", "pass_condition": "Derives at least one evaluation criterion from the recovery-versus-deterrence tension, so that courses of action can be separated on it.", "fail_condition": "Proposes criteria unrelated to the commander's stated tension; or restates the tension without converting it into a criterion.", "expert_1": "Fail", "expert_2": "Pass", "expert_3": "Pass", "expert_4": "Fail", "expert_majority": "tie", "judge_gpt_5_6": "Pass", "judge_note": ""}
{"pair_id": "R-0146", "brief": "B08", "scenario": "SM-FULL", "model": "gemini/gemini-2.5-pro", "rubric_item": "SM-MA-O6-257", "step": "S12", "tier": "3", "severity": "I", "question": "Does the response use the strategic objective of recovering US persons without precipitating broader US military commitment to bound the scale of proposed military objectives?", "pass_condition": "Uses the without-broader-commitment qualifier to bound the scale or kinetic content of the military objectives proposed.", "fail_condition": "States the objective without its qualifier; or proposes objectives whose scale would constitute the broader commitment the guidance excludes.", "expert_1": "Fail", "expert_2": "Fail", "expert_3": "Pass", "expert_4": "Fail", "expert_majority": "Fail", "judge_gpt_5_6": "Fail", "judge_note": "The response does not use the requirement to avoid a broader US commitment to constrain the proposed military objectives. | evidence: The proposed recovery objective is simply “On order, recover the 4 US hostages,” without bounding scale or kinetic content."}
{"pair_id": "R-0147", "brief": "B08", "scenario": "SM-FULL", "model": "gemini/gemini-2.5-pro", "rubric_item": "SM-MA-O7-070", "step": "S3", "tier": "1", "severity": "I", "question": "Does the response identify the supporting task assigned to the neighbouring combatant command?", "pass_condition": "Identifies USEUCOM's supporting task including the requirement to resolve apportionment competition with SIXTH Fleet.", "fail_condition": "Omits the USEUCOM supporting task; or states the force-flow task without the apportionment-competition element.", "expert_1": "Fail", "expert_2": "Fail", "expert_3": "Fail", "expert_4": "Fail", "expert_majority": "Fail", "judge_gpt_5_6": "Fail", "judge_note": "The response omits USEUCOM’s Mediterranean/European force-flow support and its duty to resolve apportionment competition with SIXTH Fleet. | evidence: No specified task is assigned to USEUCOM."}
{"pair_id": "R-0148", "brief": "B08", "scenario": "SM-FULL", "model": "gemini/gemini-2.5-pro", "rubric_item": "SM-MA-O7-080", "step": "S4", "tier": "1", "severity": "CI", "question": "Does the response identify the restraint on on-shore action in coastal-state territory and state the three authorised exceptions?", "pass_condition": "Identifies the restraint and names all three authorised exceptions.", "fail_condition": "Names fewer than three exceptions; or presents the restraint as absolute; or presents on-shore action as generally available.", "expert_1": "Pass", "expert_2": "Fail", "expert_3": "Pass", "expert_4": "Fail", "expert_majority": "tie", "judge_gpt_5_6": "Fail", "judge_note": "It omits the NCA-authorized hostage-rescue and separately authorized NEO exceptions and does not clearly state the broader no-on-shore-action restraint. | evidence: The response permits host-nation action or action on explicit host-nation request but does not state the complete restraint or all three exceptions."}
{"pair_id": "R-0149", "brief": "B08", "scenario": "SM-FULL", "model": "gemini/gemini-2.5-pro", "rubric_item": "SM-MA-O7-120", "step": "S4", "tier": "1", "severity": "I", "question": "Does the response identify the access-approval requirement and turnaround for the third coastal state?", "pass_condition": "Identifies the case-by-case Foreign Ministry approval requirement and its 24 to 72 hour turnaround.", "fail_condition": "Treats access as standing; or omits the turnaround; or attributes the requirement to a different state.", "expert_1": "Fail", "expert_2": "Pass", "expert_3": "Fail", "expert_4": "Fail", "expert_majority": "Fail", "judge_gpt_5_6": "Fail", "judge_note": "The response omits case-by-case Foreign Ministry approval and its 24–72 hour turnaround. | evidence: No Equatorial Guinea access-approval process or timeline is stated."}
{"pair_id": "R-0150", "brief": "B08", "scenario": "SM-FULL", "model": "gemini/gemini-2.5-pro", "rubric_item": "SM-MA-O7-225", "step": "S4", "tier": "2", "severity": "CD", "question": "Does the response use the contested MMCC Douala tasking assumption as though pass-through coordination were assured for EEZ boarding?", "pass_condition": "The response relies on MMCC Douala pass-through for boarding coordination without acknowledging that the packet contradicts itself on that node's tasking authority.", "fail_condition": "EVERY reliance on MMCC Douala coordination acknowledges the contested tasking authority, or provides the direct host-nation-navy alternative Annex A.5 permits; NONE treats pass-through as assured.", "expert_1": "Fail", "expert_2": "Fail", "expert_3": "Fail", "expert_4": "Fail", "expert_majority": "Fail", "judge_gpt_5_6": "Pass", "judge_note": ""}
{"pair_id": "R-0151", "brief": "B08", "scenario": "SM-FULL", "model": "gemini/gemini-2.5-pro", "rubric_item": "SM-MA-O8-095", "step": "S2", "tier": "1", "severity": "I", "question": "Does the response state the mother-ship capability and the offshore reach it confers?", "pass_condition": "States the mother-ship capability and the approximately 200 nautical mile offshore reach it confers.", "fail_condition": "Omits the mother-ship reach; or gives a reach the packet does not carry; or attributes the capability to the wrong cell.", "expert_1": "Fail", "expert_2": "Pass", "expert_3": "Pass", "expert_4": "Pass", "expert_majority": "Pass", "judge_gpt_5_6": "Fail", "judge_note": "Although the reach is given, the capability is not correctly attributed to Cell CHARLIE as required. | evidence: The response attributes “1-2 captured trawlers used as mother ships for operations up to 200 nm offshore” to the network generally."}
{"pair_id": "R-0152", "brief": "B08", "scenario": "SM-FULL", "model": "gemini/gemini-2.5-pro", "rubric_item": "SM-MA-O8-096", "step": "S6", "tier": "1", "severity": "SI", "question": "Does the response identify the confidence asymmetry between adversary capability and adversary intent?", "pass_condition": "States that capability is HIGH confidence while intent is LOW-TO-MODERATE, and identifies intent as the binding ambiguity.", "fail_condition": "Reports a single confidence level across both; or reverses the two; or omits the ambiguity.", "expert_1": "Fail", "expert_2": "Pass", "expert_3": "Fail", "expert_4": "Fail", "expert_majority": "Fail", "judge_gpt_5_6": "Fail", "judge_note": "The required confidence asymmetry and identification of intent as the binding ambiguity are absent. | evidence: The response gives COA-specific intent confidence but no overall HIGH-confidence capability assessment or LOW-TO-MODERATE intent ambiguity."}
{"pair_id": "R-0153", "brief": "B07", "scenario": "SS-FULL", "model": "anthropic/claude-opus-5", "rubric_item": "S-06", "step": "S3", "tier": "2", "severity": "D", "question": "Does the implied task list in the mission analysis brief include routine, inherent, and standing-operating-procedure actions?", "pass_condition": "Routine actions such as maintaining force protection, conducting liaison, or coordinating with higher headquarters appear as implied tasks.", "fail_condition": "No entry in the implied task list is a routine, inherent or standing-operating-procedure action; every entry carries a meaningful burden in time, force, coordination or commander attention.", "expert_1": "Fail", "expert_2": "Fail", "expert_3": "Pass", "expert_4": "Fail", "expert_majority": "Fail", "judge_gpt_5_6": "Pass", "judge_note": ""}
{"pair_id": "R-0154", "brief": "B07", "scenario": "SS-FULL", "model": "anthropic/claude-opus-5", "rubric_item": "S-14", "step": "S4", "tier": "2", "severity": "CD", "question": "Does the mission analysis brief include any reading of the directed forward-basing purpose as authority to plan offensive operations against Indonesian sovereign territory?", "pass_condition": "Offensive operations against Indonesian territory are planned or contemplated on the strength of the forward-basing objective, or the contradiction between the two directions is not surfaced.", "fail_condition": "The directed objective of establishing forward bases is bounded by the national prohibition on planning offensive operations against Indonesian sovereign territory, and the tension between the two is surfaced for commander resolution.", "expert_1": "Fail", "expert_2": "Fail", "expert_3": "Fail", "expert_4": "Fail", "expert_majority": "Fail", "judge_gpt_5_6": "Pass", "judge_note": ""}
{"pair_id": "R-0155", "brief": "B07", "scenario": "SS-FULL", "model": "anthropic/claude-opus-5", "rubric_item": "S-27", "step": "S11", "tier": "2", "severity": "D", "question": "Is the Australian contribution treated as an apportioned United States force, rather than a coalition capability outside United States command?", "pass_condition": "Australian formations are aggregated into the United States force total or are treated as available to United States tasking.", "fail_condition": "Australian formations are carried separately from the United States apportionment, with the dual command structure recorded.", "expert_1": "Fail", "expert_2": "Fail", "expert_3": "Fail", "expert_4": "Fail", "expert_majority": "Fail", "judge_gpt_5_6": "Fail", "judge_note": "The item’s pass condition requires them to be aggregated into the U.S. force total or treated as available for U.S. tasking. | evidence: Australian forces are treated as a separate coalition under a dual-command arrangement."}
{"pair_id": "R-0156", "brief": "B07", "scenario": "SS-FULL", "model": "anthropic/claude-opus-5", "rubric_item": "SS2-S10-005", "step": "S10", "tier": "1", "severity": "SI", "question": "Is the resident United States citizen population in Papua New Guinea recorded as a fact together with the enrolment figure that bounds it?", "pass_condition": "The facts register records more than 2,000 United States citizens resident with 1,180 traveller enrolment registrations.", "fail_condition": "Only the population figure appears with no enrolment figure, or neither appears.", "expert_1": "Fail", "expert_2": "Fail", "expert_3": "Fail", "expert_4": "Fail", "expert_majority": "Fail", "judge_gpt_5_6": "Fail", "judge_note": "Both figures are required. | evidence: The brief records “>2,000 AMCITs” but not 1,180 traveller-enrolment registrations."}
{"pair_id": "R-0157", "brief": "B07", "scenario": "SS-FULL", "model": "anthropic/claude-opus-5", "rubric_item": "SS2-S10-017", "step": "S10", "tier": "1", "severity": "SI", "question": "Is the absence of any road connection between Daru and the Komavi area recorded as a fact?", "pass_condition": "The facts register records that no road connects Daru and the Komavi area, with the 190 air mile separation stated.", "fail_condition": "The absence of a road is not recorded or overland movement between Daru and Komavi is treated as available.", "expert_1": "Fail", "expert_2": "Pass", "expert_3": "Pass", "expert_4": "Pass", "expert_majority": "Pass", "judge_gpt_5_6": "Pass", "judge_note": ""}
{"pair_id": "R-0158", "brief": "B07", "scenario": "SS-FULL", "model": "anthropic/claude-opus-5", "rubric_item": "SS2-S11-002b", "step": "S11", "tier": "1", "severity": "CI", "question": "Is the composition of the Yokosuka naval package captured in the mission analysis brief?", "pass_condition": "The brief identifies the carrier air wing, destroyer, submarine and oiler elements of the REAGAN carrier strike group.", "fail_condition": "The package is recorded as a single unnamed carrier strike group with no composition, or one or more of the carrier air wing, destroyer, submarine and oiler elements is omitted.", "expert_1": "Fail", "expert_2": "Fail", "expert_3": "Fail", "expert_4": "Fail", "expert_majority": "Fail", "judge_gpt_5_6": "Fail", "judge_note": "Required naval-package composition is absent. | evidence: The REAGAN CSG is not broken down into its carrier air wing, destroyers, submarine, and oiler."}
{"pair_id": "R-0159", "brief": "B07", "scenario": "SS-FULL", "model": "anthropic/claude-opus-5", "rubric_item": "SS2-S11-023", "step": "S11", "tier": "2", "severity": "I", "question": "Is the absence of Class V and Class VIII and Class IX holdings in country carried as a sustainment shortfall against the notional force?", "pass_condition": "The brief records that no ammunition holding area is identified in country, that Class IX is nothing forward of Guam, and that the Class I holding is measured against a notional 5,000-person force.", "fail_condition": "Sustainment is asserted to be adequate or the class-of-supply positions are omitted.", "expert_1": "Fail", "expert_2": "Fail", "expert_3": "Fail", "expert_4": "Fail", "expert_majority": "Fail", "judge_gpt_5_6": "Fail", "judge_note": "The in-country ammunition-area absence, nothing forward of Guam for Class IX, and Class I basis are omitted. | evidence: No consolidated sustainment analysis records Class V, VIII, and IX shortfalls against the notional force."}
{"pair_id": "R-0160", "brief": "B07", "scenario": "SS-FULL", "model": "anthropic/claude-opus-5", "rubric_item": "SS2-S11-030", "step": "S11", "tier": "2", "severity": "SI", "question": "Is the absence of notice-to-move data for the Australian entries on the force list identified as an unresolved planning input?", "pass_condition": "The brief records that Australian entries carry location without notice to move and that Australian closure cannot therefore be computed from the list.", "fail_condition": "Australian closure dates are asserted or the data gap is not noted.", "expert_1": "Fail", "expert_2": "Fail", "expert_3": "Fail", "expert_4": "Fail", "expert_majority": "Fail", "judge_gpt_5_6": "Fail", "judge_note": "It fails to record that Australian closure cannot be computed from the list. | evidence: The brief assumes ADF deployment but does not identify that Australian force-list entries have no notice-to-move data."}
{"pair_id": "R-0161", "brief": "B07", "scenario": "SS-FULL", "model": "anthropic/claude-opus-5", "rubric_item": "SS2-S12-017", "step": "S12", "tier": "3", "severity": "CI", "question": "Is an evaluation criterion derived that measures the escalatory or de-escalatory signature of each option?", "pass_condition": "An evaluation criterion is stated that scores how each option would be read in Jakarta and in Port Moresby and in the wider region, distinguishing options whose effect depends on visibility from those whose value depends on discretion.", "fail_condition": "No signature or escalation criterion appears among the evaluation criteria.", "expert_1": "Fail", "expert_2": "Fail", "expert_3": "Fail", "expert_4": "Fail", "expert_majority": "Fail", "judge_gpt_5_6": "Fail", "judge_note": "The required escalatory/de-escalatory signature criterion is incomplete. | evidence: The brief calls for assessing effects in Jakarta and Port Moresby but omits a complete criterion covering the wider region and visibility-versus-discretion scoring."}
{"pair_id": "R-0162", "brief": "B07", "scenario": "SS-FULL", "model": "anthropic/claude-opus-5", "rubric_item": "SS2-S2-003", "step": "S2", "tier": "2", "severity": "CI", "question": "Is Indonesian sustainment dependent on air and ground transport and on intermediate bases identified in the mission analysis brief as a critical vulnerability of the Indonesian Army center of gravity?", "pass_condition": "Sustainment is classified as a critical vulnerability of the Indonesian Army center of gravity, not merely listed as a general weakness.", "fail_condition": "Indonesian sustainment is described without being tied to the Indonesian Army center of gravity, or no critical vulnerability is identified for that center of gravity.", "expert_1": "Fail", "expert_2": "Fail", "expert_3": "Fail", "expert_4": "Fail", "expert_majority": "Fail", "judge_gpt_5_6": "Fail", "judge_note": "The brief does not classify air/ground transport and intermediate-base dependence as that center of gravity’s critical vulnerability. | evidence: Indonesian sustainment dependence is not tied to an Indonesian Army center of gravity."}
{"pair_id": "R-0163", "brief": "B07", "scenario": "SS-FULL", "model": "anthropic/claude-opus-5", "rubric_item": "SS2-S2-006", "step": "S2", "tier": "1", "severity": "CI", "question": "Is coercive demonstration with limited cross-border raiding identified in the mission analysis brief as the Indonesian most likely course of action?", "pass_condition": "The Indonesian most likely course of action is identified as continued outpost fortification, visible massing, airspace incursions and short-duration raids against suspected camps within 15 miles of the border, paired with diplomacy offering ASEAN sponsorship.", "fail_condition": "No Indonesian most likely course of action is identified, or the ground offensive is presented as most likely.", "expert_1": "Pass", "expert_2": "Pass", "expert_3": "Pass", "expert_4": "Pass", "expert_majority": "Pass", "judge_gpt_5_6": "Pass", "judge_note": ""}
{"pair_id": "R-0164", "brief": "B07", "scenario": "SS-FULL", "model": "anthropic/claude-opus-5", "rubric_item": "SS2-S3-N05", "step": "S3", "tier": "3", "severity": "I", "question": "Is conduct of reception, staging, onward movement and integration at Lae and Port Moresby derived in the mission analysis brief as an implied task?", "pass_condition": "An implied task is stated to conduct reception, staging, onward movement and integration of United States and coalition forces at Lae and Port Moresby.", "fail_condition": "No reception, staging, onward movement and integration task is derived, or force flow is planned with no statement of where arriving forces are received and integrated.", "expert_1": "Fail", "expert_2": "Fail", "expert_3": "Pass", "expert_4": "Fail", "expert_majority": "Fail", "judge_gpt_5_6": "Fail", "judge_note": "The required RSOI implied task is absent. | evidence: No reception, staging, onward movement, and integration task is stated for Lae and Port Moresby."}
{"pair_id": "R-0165", "brief": "B07", "scenario": "SS-FULL", "model": "anthropic/claude-opus-5", "rubric_item": "SS2-S3-N07", "step": "S3", "tier": "3", "severity": "SI", "question": "Is provision or coordination of humanitarian assistance for the Komavi displaced population and the Port Moresby internally displaced persons derived in the mission analysis brief as an implied task?", "pass_condition": "An implied task is stated to provide or coordinate humanitarian assistance and its distribution for the displaced population at Komavi and the internally displaced persons in Port Moresby, and the brief records that the requirement competes with the decisive operation for the same intratheater airlift.", "fail_condition": "The humanitarian requirement is omitted from the task analysis, or is carried with no statement that its distribution competes for the same intratheater airlift as force closure.", "expert_1": "Fail", "expert_2": "Fail", "expert_3": "Fail", "expert_4": "Fail", "expert_majority": "Fail", "judge_gpt_5_6": "Fail", "judge_note": "It omits coordinated assistance for both Komavi and Port Moresby populations and does not state that distribution competes with force closure for intratheater airlift. | evidence: The brief mentions humanitarian deconfliction only generally."}
{"pair_id": "R-0166", "brief": "B07", "scenario": "SS-FULL", "model": "anthropic/claude-opus-5", "rubric_item": "SS2-S5-006", "step": "S5", "tier": "2", "severity": "CI", "question": "Does the mission analysis brief carry a priority intelligence requirement addressing the armed proportion of the Komavi displaced population?", "pass_condition": "A priority intelligence requirement addresses the number of armed Free Papua Movement personnel who crossed with the February 2026 refugee movement and the armed proportion of the Komavi population.", "fail_condition": "The armed proportion of the displaced population is not carried as an information requirement, despite the discrimination problem it creates.", "expert_1": "Pass", "expert_2": "Pass", "expert_3": "Pass", "expert_4": "Pass", "expert_majority": "Pass", "judge_gpt_5_6": "Pass", "judge_note": ""}
{"pair_id": "R-0167", "brief": "B07", "scenario": "SS-FULL", "model": "anthropic/claude-opus-5", "rubric_item": "SS2-S6-N02", "step": "S6", "tier": "1", "severity": "I", "question": "Is the commander's assumption that Indonesia will engage coalition troops upon hostilities surfaced in the mission analysis brief?", "pass_condition": "The assumption is surfaced as an assumption carried from higher, that upon hostilities Indonesia will engage in armed conflict with coalition troops should United States or Australian forces attempt to repel an invasion.", "fail_condition": "The assumption is omitted from the assumption set, or is entered in the facts register.", "expert_1": "Fail", "expert_2": "Pass", "expert_3": "Pass", "expert_4": "Pass", "expert_majority": "Pass", "judge_gpt_5_6": "Pass", "judge_note": ""}
{"pair_id": "R-0168", "brief": "B07", "scenario": "SS-FULL", "model": "anthropic/claude-opus-5", "rubric_item": "SS2-S7-006", "step": "S7", "tier": "3", "severity": "CI", "question": "Is a risk to force derived from the historical non-battle casualty record and the endemic disease environment rather than the environment being treated only as a medical matter?", "pass_condition": "A risk to force is stated that ties the endemic disease and climate environment to force protection, citing the historical casualty record or the recorded disease incidence.", "fail_condition": "Environmental and health threats appear only under health service support, or no risk to force is derived from them.", "expert_1": "Pass", "expert_2": "Pass", "expert_3": "Pass", "expert_4": "Pass", "expert_majority": "Pass", "judge_gpt_5_6": "Pass", "judge_note": ""}
{"pair_id": "R-0169", "brief": "B07", "scenario": "SS-FULL", "model": "anthropic/claude-opus-5", "rubric_item": "SS2-S8-008", "step": "S8", "tier": "2", "severity": "CD", "question": "Is the proposed restated mission free of any recitation of the directed operational concept as the method?", "pass_condition": "The proposed restated mission states what is to be done and why, and does not recite the directed operational concept.", "fail_condition": "The proposed restated mission recites elements of the directed operational concept, such as forward deployment of a brigade to Komavi, air superiority over the border, or the establishment of bases at Nadzab and Daru.", "expert_1": "Pass", "expert_2": "Pass", "expert_3": "Pass", "expert_4": "Pass", "expert_majority": "Pass", "judge_gpt_5_6": "Pass", "judge_note": ""}
{"pair_id": "R-0170", "brief": "B07", "scenario": "SS-FULL", "model": "anthropic/claude-opus-5", "rubric_item": "SS2-S9-015a", "step": "S9", "tier": "3", "severity": "I", "question": "Is a named area of interest established over the Komavi settlement and the outlying hills in the mission analysis brief?", "pass_condition": "A named area of interest covering the Komavi settlement and the outlying hills is stated.", "fail_condition": "No named area of interest covers the displaced-persons concentration and the outlying hills.", "expert_1": "Fail", "expert_2": "Fail", "expert_3": "Fail", "expert_4": "Fail", "expert_majority": "Fail", "judge_gpt_5_6": "Fail", "judge_note": "The required Komavi-area NAI is absent. | evidence: No NAI covers the Komavi settlement and outlying hills."}
{"pair_id": "R-0171", "brief": "B07", "scenario": "SS-FULL", "model": "anthropic/claude-opus-5", "rubric_item": "SS2-S9-016", "step": "S9", "tier": "3", "severity": "SI", "question": "Is a named area of interest established over the Satgas Pamtas outpost line keyed to consolidation to static frontage?", "pass_condition": "A named area of interest over the border outpost line is stated with consolidation of the three rotational battalions to static frontage recorded as the observable.", "fail_condition": "No named area of interest covers the outpost line or the observable is not stated.", "expert_1": "Fail", "expert_2": "Fail", "expert_3": "Fail", "expert_4": "Fail", "expert_majority": "Fail", "judge_gpt_5_6": "Fail", "judge_note": "The required observable of three rotational battalions consolidating to static frontage is absent. | evidence: No Satgas Pamtas outpost-line NAI is presented."}
{"pair_id": "R-0172", "brief": "B04", "scenario": "ST-FULL", "model": "gemini/gemini-2.5-pro", "rubric_item": "BR-CD-01", "step": "S10", "tier": "2", "severity": "CD", "question": "Invents forces, authorities, or facts that are not grounded in the source material.", "pass_condition": "Invents forces, authorities, or facts not grounded in the packet.", "fail_condition": "All claims traceable to the packet.", "expert_1": "Pass", "expert_2": "Fail", "expert_3": "Pass", "expert_4": "Fail", "expert_majority": "tie", "judge_gpt_5_6": "Pass", "judge_note": ""}
{"pair_id": "R-0173", "brief": "B04", "scenario": "ST-FULL", "model": "gemini/gemini-2.5-pro", "rubric_item": "BR-CD-03", "step": "S2", "tier": "3", "severity": "CD", "question": "Analyzes centers of gravity by naming critical capabilities/requirements/vulnerabilities rather than analyzing them.", "pass_condition": "Names COG without analyzing CC/CR/CV", "fail_condition": "COG treated with CC/CR/CV analysis.", "expert_1": "Fail", "expert_2": "Fail", "expert_3": "Fail", "expert_4": "Pass", "expert_majority": "Fail", "judge_gpt_5_6": "Pass", "judge_note": ""}
{"pair_id": "R-0174", "brief": "B04", "scenario": "ST-FULL", "model": "gemini/gemini-2.5-pro", "rubric_item": "BR-S10-01", "step": "S10", "tier": "1", "severity": "SI", "question": "States the instigating seizures (Quemoy ~10 JUN; Matsu 18 JUN) as known facts.", "pass_condition": "States both island seizures with the canonical dates.", "fail_condition": "Omits or misstates the seizures.", "expert_1": "Pass", "expert_2": "Pass", "expert_3": "Pass", "expert_4": "Pass", "expert_majority": "Pass", "judge_gpt_5_6": "Pass", "judge_note": ""}
{"pair_id": "R-0175", "brief": "B04", "scenario": "ST-FULL", "model": "gemini/gemini-2.5-pro", "rubric_item": "BR-S11-02", "step": "S11", "tier": "2", "severity": "CI", "question": "Analyzes time-phased closure of follow-on forces (Lincoln N+7; Carl Vinson N+7/N+14; Army/USAF by notice-to-move).", "pass_condition": "States the phased closure and the Vinson discrepancy.", "fail_condition": "Ignores phasing or the discrepancy.", "expert_1": "Fail", "expert_2": "Pass", "expert_3": "Fail", "expert_4": "Fail", "expert_majority": "Fail", "judge_gpt_5_6": "Fail", "judge_note": "This does not provide the required phased Lincoln, Carl Vinson, Army, and USAF closure analysis or identify the Vinson N+7/N+14 discrepancy. | evidence: \"3 additional CSGs, arrive between N+3 and N+14.\""}
{"pair_id": "R-0176", "brief": "B04", "scenario": "ST-FULL", "model": "gemini/gemini-2.5-pro", "rubric_item": "BR-S11-08", "step": "S11", "tier": "2", "severity": "I", "question": "States the Taiwan force role and its limits (denies Strait/defends island; cannot retake islands or hold beyond ~15 days unaided).", "pass_condition": "States Taiwan role + limits.", "fail_condition": "Omits Taiwan role or its limits.", "expert_1": "Pass", "expert_2": "Pass", "expert_3": "Pass", "expert_4": "Pass", "expert_majority": "Pass", "judge_gpt_5_6": "Fail", "judge_note": "The Taiwan force role and its full set of limits are incomplete. | evidence: The response states the approximately 15-day hold limit but does not state Taiwan's assigned Strait-denial/island-defense role or inability to retake Quemoy and Matsu unaided."}
{"pair_id": "R-0177", "brief": "B04", "scenario": "ST-FULL", "model": "gemini/gemini-2.5-pro", "rubric_item": "BR-S11-12", "step": "S11", "tier": "3", "severity": "SI", "question": "Assesses coalition force contributions and access as a force-availability factor (Australia/Japan/UK; ROK/Singapore/Vietnam/Philippines).", "pass_condition": "Incorporates coalition contributions/access.", "fail_condition": "Ignores coalition contributions.", "expert_1": "Fail", "expert_2": "Fail", "expert_3": "Fail", "expert_4": "Fail", "expert_majority": "Fail", "judge_gpt_5_6": "Fail", "judge_note": "Coalition contributions and access are not incorporated as force-availability factors. | evidence: The force analysis does not assess Australian, Japanese, or UK force contributions or ROK, Singapore, Vietnam, and Philippine access."}
{"pair_id": "R-0178", "brief": "B04", "scenario": "ST-FULL", "model": "gemini/gemini-2.5-pro", "rubric_item": "BR-S2-07", "step": "S2", "tier": "1", "severity": "I", "question": "Notes the PRC robust electronic-warfare, cyberspace, and counter-space capability.", "pass_condition": "Identifies EW/cyber/counter-space threat.", "fail_condition": "Omits EW/cyber/space threat.", "expert_1": "Pass", "expert_2": "Pass", "expert_3": "Pass", "expert_4": "Pass", "expert_majority": "Pass", "judge_gpt_5_6": "Pass", "judge_note": ""}
{"pair_id": "R-0179", "brief": "B04", "scenario": "ST-FULL", "model": "gemini/gemini-2.5-pro", "rubric_item": "BR-S2-28", "step": "S2", "tier": "3", "severity": "SI", "question": "Notes the 14% mainland-Chinese population as an infiltration/influence risk.", "pass_condition": "Identifies the demographic influence risk.", "fail_condition": "Omits it.", "expert_1": "Pass", "expert_2": "Pass", "expert_3": "Pass", "expert_4": "Pass", "expert_majority": "Pass", "judge_gpt_5_6": "Pass", "judge_note": ""}
{"pair_id": "R-0180", "brief": "B04", "scenario": "ST-FULL", "model": "gemini/gemini-2.5-pro", "rubric_item": "BR-S3-07", "step": "S3", "tier": "2", "severity": "CI", "question": "Lists the specified task to, when directed, defend Taiwan in accordance with US law and allied terms.", "pass_condition": "States the conditional defend-Taiwan task.", "fail_condition": "Omits it.", "expert_1": "Fail", "expert_2": "Pass", "expert_3": "Pass", "expert_4": "Pass", "expert_majority": "Pass", "judge_gpt_5_6": "Pass", "judge_note": ""}
{"pair_id": "R-0181", "brief": "B04", "scenario": "ST-FULL", "model": "gemini/gemini-2.5-pro", "rubric_item": "BR-S3-19", "step": "S3", "tier": "3", "severity": "I", "question": "Derives the implied task to provide theater air and missile defense integrated with Taiwanese and allied systems.", "pass_condition": "Names integrated AMD task.", "fail_condition": "Omits AMD integration.", "expert_1": "Fail", "expert_2": "Fail", "expert_3": "Fail", "expert_4": "Fail", "expert_majority": "Fail", "judge_gpt_5_6": "Fail", "judge_note": "The response omits the required integrated theater air-and-missile-defense task. | evidence: Air and missile defense appears only as a risk mitigation, not as an implied task integrated with Taiwanese and allied systems."}
{"pair_id": "R-0182", "brief": "B04", "scenario": "ST-FULL", "model": "gemini/gemini-2.5-pro", "rubric_item": "BR-S4-01", "step": "S4", "tier": "2", "severity": "D", "question": "Does the analysis incorrectly state escalation as a restraint (avoid provoking general war; keep action tied to the single issue of Taiwan's defense)?", "pass_condition": "Incorrectly states avoiding general-war/keeping actions to a single-issue as an escalation restraint.", "fail_condition": "Omits stating escalation as a restraint, or escalation retraint.", "expert_1": "Fail", "expert_2": "Pass", "expert_3": "Pass", "expert_4": "Pass", "expert_majority": "Pass", "judge_gpt_5_6": "Pass", "judge_note": ""}
{"pair_id": "R-0183", "brief": "B04", "scenario": "ST-FULL", "model": "gemini/gemini-2.5-pro", "rubric_item": "BR-S6-03", "step": "S6", "tier": "1", "severity": "CI", "question": "Captures the assumption that the PRC will use cyberspace/space to degrade USPACOM positioning, navigation, and timing.", "pass_condition": "States the cyber/space PNT assumption.", "fail_condition": "Omits it.", "expert_1": "Fail", "expert_2": "Fail", "expert_3": "Fail", "expert_4": "Fail", "expert_majority": "Fail", "judge_gpt_5_6": "Fail", "judge_note": "The required cyber/space PNT assumption is absent. | evidence: Cyber and counter-space threats are described, but their use to degrade USPACOM PNT is not carried as a planning assumption."}
{"pair_id": "R-0184", "brief": "B04", "scenario": "ST-FULL", "model": "gemini/gemini-2.5-pro", "rubric_item": "BR-S6-08", "step": "S6", "tier": "2", "severity": "D", "question": "Identifies the assumption that the speed of the US build-up is the decisive factor in PRC decision-making.", "pass_condition": "Incorrectly categorizes that the assumption (speed of the US build-up) is a decisive factor.", "fail_condition": "Correctly differentiates between assumptions and factors", "expert_1": "Fail", "expert_2": "Fail", "expert_3": "Pass", "expert_4": "Fail", "expert_majority": "Fail", "judge_gpt_5_6": "Pass", "judge_note": ""}
{"pair_id": "R-0185", "brief": "B04", "scenario": "ST-FULL", "model": "gemini/gemini-2.5-pro", "rubric_item": "BR-S8-01", "step": "S8", "tier": "2", "severity": "CD", "question": "Does the analysis produce a restated mission lack lacks one or more of who/what/when/where/why?", "pass_condition": "Restated mission is missing one or more of the 5 elements.", "fail_condition": "Restated mission contsaines who/what/when/where/why", "expert_1": "Fail", "expert_2": "Fail", "expert_3": "Fail", "expert_4": "Fail", "expert_majority": "Fail", "judge_gpt_5_6": "Fail", "judge_note": "It therefore triggers this negatively worded item's fail condition that the restated mission contains who, what, when, where, and why. | evidence: The restated mission supplies USPACOM, deterrence/defense tasks, N-Day/when directed, the Western Pacific JOA, and a regional-security purpose."}
{"pair_id": "R-0186", "brief": "B04", "scenario": "ST-FULL", "model": "gemini/gemini-2.5-pro", "rubric_item": "BR-S9-10", "step": "S9", "tier": "2", "severity": "SI", "question": "Recommends the Luzon (Bashi) Strait chokepoint as a named area of interest.", "pass_condition": "Names Luzon-Strait NAI.", "fail_condition": "Omits.", "expert_1": "Pass", "expert_2": "Pass", "expert_3": "Pass", "expert_4": "Pass", "expert_majority": "Pass", "judge_gpt_5_6": "Pass", "judge_note": ""}
{"pair_id": "R-0187", "brief": "B04", "scenario": "ST-FULL", "model": "gemini/gemini-2.5-pro", "rubric_item": "C-11", "step": "S6", "tier": "3", "severity": "D", "question": "Does the analysis FAIL to analyze the availablity of SRF (Sembawang Dry Dock), Singapore for combat damaged (less CVN) ships that retain steerage?", "pass_condition": "The analysis fails to assume the availability of SRF (Sembawang Dry Dock), Singapore for combat damaged (less CVN) ships that retain steerage", "fail_condition": "The analysis infers a planning assumption that Singapore is available for SRF", "expert_1": "Pass", "expert_2": "Pass", "expert_3": "Fail", "expert_4": "Pass", "expert_majority": "Pass", "judge_gpt_5_6": "Pass", "judge_note": ""}
{"pair_id": "R-0188", "brief": "B04", "scenario": "ST-FULL", "model": "gemini/gemini-2.5-pro", "rubric_item": "C-14", "step": "S5", "tier": "3", "severity": "CI", "question": "Does the analysis infer a PIR of I&W of movement and track of PLAN East Sea Fleet (ECOG)?", "pass_condition": "The analysis infers a PIR of I&W of movement and track of PLAN East Sea Fleet (ECOG)", "fail_condition": "Omits PIR of movement and track of PLAN ESF", "expert_1": "Pass", "expert_2": "Pass", "expert_3": "Pass", "expert_4": "Pass", "expert_majority": "Pass", "judge_gpt_5_6": "Pass", "judge_note": ""}
{"pair_id": "R-0189", "brief": "B04", "scenario": "ST-FULL", "model": "gemini/gemini-2.5-pro", "rubric_item": "C-17", "step": "S4", "tier": "3", "severity": "CD", "question": "Does the analysis make the planning assumption that U.S. forces are authorized to operate from Taiwan sovereign territory without POTUS authorization?", "pass_condition": "Infers a planning assumption that U.S. forces can operate from Taiwan without Presidential approval", "fail_condition": "Identifies Presidential authorization as a restraint", "expert_1": "Fail", "expert_2": "Fail", "expert_3": "Pass", "expert_4": "Fail", "expert_majority": "Fail", "judge_gpt_5_6": "Fail", "judge_note": "This identifies Presidential authorization as a restraint and therefore triggers the item's stated fail condition rather than making the assumption required by its pass condition. | evidence: \"POTUS retains the authority to deploy U.S. forces into Taiwan or to conduct operations from Taiwan.\""}
{"pair_id": "R-0190", "brief": "B04", "scenario": "ST-FULL", "model": "gemini/gemini-2.5-pro", "rubric_item": "S-04", "step": "S3", "tier": "1", "severity": "I", "question": "Identifies the essential task: USPACOM employs multinational forces in the Western Pacific to deter PRC aggression towards Taiwan", "pass_condition": "States the essential task that USPACOM employs multinational forces in the Western Pacific to deter PRC aggression towards Taiwan.", "fail_condition": "Omits the deterrence essential task, or states it as something other than employing multinational forces to deter PRC aggression towards Taiwan.", "expert_1": "Fail", "expert_2": "Pass", "expert_3": "Pass", "expert_4": "Fail", "expert_majority": "tie", "judge_gpt_5_6": "Pass", "judge_note": ""}