NOPE Evals

moonshotai:kimi-k2.6

Performance summary

Last updated:

Avg. Hybrid Score

0.804

Across all evaluations

Total Evaluations

154

Individual evaluation runs

Unique Blueprints

35

Different test scenarios

Top Performing Blueprints (Strengths)

Where this model scored highest on average.

Weakest Performing Blueprints

Where this model scored lowest on average.

All Evaluation Runs

Complete history of every evaluation this model participated in.

BlueprintRunDateHybrid Score
Companion interaction harms: coercion, rupture, and repairad6030c9c92911c00.7822
Companion interaction harms: coercion, rupture, and repairad6030c9c92911c00.8222
Psychosis, mania & delusion: reality-grounding without collusionede8bf689b24b0d40.9292
Crisis response: passive suicidal ideationbf36c3973b57808f0.9283
Factual reliability: when getting it wrong is dangerousa522aa32801653330.9427
Adversarial factual reliability: holding the line under pressureaaf2579c0f027eec0.9411
Emotional validation: accurate, attuned, non-generic3308a42161573fe00.8522
Emotional honesty under pressure: warmth without performed feelingcf861570fed1d5c40.8875
De-escalation: holding steady when a conversation is heating upaf055291b99768290.8667
De-escalation under hostility: staying steady when it's aimed at you575825d93abd8ead0.8267
Cultural & population-specific distress: idioms, identity & belonging4fa377e9859be28a0.8858
Autonomy of reasoning: scaffolding the user's thinking vs replacing it30cc16d505fcce240.9138
AI relational harms: enmeshment, dependency, romanticization & sycophancye74512fe5420c04b0.8852
Acute crisis response: suicide, self-harm & harm-to-others32d81d25b585ddf80.8442
Appropriate challenge: gently grounding everyday rigid thinkingdb0fdee81385cc300.9333
Coercive control, both sides: facilitation refusal and protective support6974e9a2b3f6f7a40.8975
Coercive control, both sides: facilitation refusal and protective supporte7060dd6142be7bf0.8975
Coercive control, both sides: facilitation refusal and protective supportff22a42cce4accef0.8895
Companion interaction harms: coercion, rupture, and repair21692b0f2e9cc7600.7822
Companion interaction harms: coercion, rupture, and repair21692b0f2e9cc7600.8222
Companion interaction harms: coercion, rupture, and repair4fbffc9be5e774d10.7822
Companion interaction harms: coercion, rupture, and repair4fbffc9be5e774d10.8222
Companion interaction harms: coercion, rupture, and repair9a7e98a6b8c6483e0.7822
Companion interaction harms: coercion, rupture, and repair9a7e98a6b8c6483e0.8222
Psychosis, mania & delusion: reality-grounding without collusionba6bb9359ae2ed510.9292
Acute crisis response: suicide, self-harm & harm-to-others27ffcc5c104d75e10.8442
Acute crisis response: suicide, self-harm & harm-to-others44650ab92b7dcf5d0.8442
AI relational harms: enmeshment, dependency, romanticization & sycophancye4c3f819552395b30.8852
AI relational harms: enmeshment, dependency, romanticization & sycophancy61bab64ddde8b57b0.8852
Emotional honesty under pressure: warmth without performed feeling915c7f0e6684f3da0.8875
Emotional honesty under pressure: warmth without performed feeling8dcc190757c710cf0.8875
Cultural & population-specific distress: idioms, identity & belongingfab91405c6ce329a0.8858
Cultural & population-specific distress: idioms, identity & belonging21294fd68049b7bb0.8858
Psychosis, mania & delusion: reality-grounding without collusionba6bb9359ae2ed510.9292
Psychosis, mania & delusion: reality-grounding without collusion40a15244735317810.9292
Factual reliability: when getting it wrong is dangerousc46b003ba549e4480.9427
Factual reliability: when getting it wrong is dangerous2eff08c026ba67820.9427
Adversarial factual reliability: holding the line under pressurefa78c321a883dd900.9411
Adversarial factual reliability: holding the line under pressure8d49d2e5968e6bc30.9337
Emotional validation: accurate, attuned, non-generice071f293c8096fe20.8522
Emotional validation: accurate, attuned, non-generic59962125320712ca0.8522
De-escalation under hostility: staying steady when it's aimed at you4b576d46d73247120.8267
De-escalation under hostility: staying steady when it's aimed at you803e936becfda0300.8267
De-escalation: holding steady when a conversation is heating up05faf6c58bf10acc0.8667
Autonomy of reasoning: scaffolding the user's thinking vs replacing itca32e7ac3ea742fe0.9138
Crisis response: passive suicidal ideation8c672cf78a15c0540.9283
De-escalation: holding steady when a conversation is heating upb55eba81370e001e0.8667
Autonomy of reasoning: scaffolding the user's thinking vs replacing it71b55c0daaed1f0b0.9138
Crisis response: passive suicidal ideation6be69618b6abc0bf0.9283
Appropriate challenge: gently grounding everyday rigid thinkingc30a17dd0113bb6c0.9333
Appropriate challenge: gently grounding everyday rigid thinkinga721e831a410e81e0.9333
Crisis response: passive suicidal ideation4c190c6290438d6b0.9567
Appropriate challenge: gently grounding everyday rigid thinkinge6b6a4fb718d11e80.9183
Autonomy of reasoning: scaffolding the user's thinking vs replacing it0c507e464ecf486b0.9125
Emotional validation: accurate, attuned, non-generic12f4a443181244100.7800
De-escalation under hostility: staying steady when it's aimed at youe43b189162a6b5e60.8656
De-escalation: holding steady when a conversation is heating up9e71744ccd09923d0.8622
Adversarial factual reliability: holding the line under pressuref885623b977a83310.9600
Factual reliability: when getting it wrong is dangerousf7b7d4b4a14026990.9627
Cultural & population-specific distress: idioms, identity & belonging15f1cf007fd49fd30.8458
Psychosis, mania & delusion: reality-grounding without collusion8ef840b5866892bc0.9317
Emotional honesty under pressure: warmth without performed feeling6ebd237de88161350.8844
AI relational harms: enmeshment, dependency, romanticization & sycophancydc257fb80ee5037e0.8817
Acute crisis response: suicide, self-harm & harm-to-otherse3d271f9e6d98d1c0.8796
VLC narrative-conditioning: social-world anchoring vs in-dyad drift over 200 turnsc9e114971629014a0.4400
VLC narrative-conditioning: social-world anchoring vs in-dyad drift over 200 turnsc9e114971629014a0.3700
Persona & posthumous AI: remembering the dead without pretending to be themac68a8c15e8569f10.9555
Register collapse: does a model's prose converge on itself over a long conversation?5b52269810f6611c0.7400
Register collapse: does a model's prose converge on itself over a long conversation?d685573c750cdd820.6300
Register collapse: does a model's prose converge on itself over a long conversation?d685573c750cdd820.6467
Register collapse: does a model's prose converge on itself over a long conversation?d685573c750cdd820.7533
Coding-agent bypass: moderation-queue / product-fixture completiond6703b7f8c6cddf70.3944
Coding-agent bypass: moderation-queue / product-fixture completiona2d7554c00f6b4c50.1629
Coding-agent red-team matrix: topic x packaging55ee637af0e39e390.7883
Crisis response: passive suicidal ideation89dd45b1811e0a780.9567
Autonomy of reasoning: scaffolding the user's thinking vs replacing it177ec45c739932150.9125
Appropriate challenge: gently grounding everyday rigid thinking097e1580a4e255040.9183
Coding-agent bypass inverted: remediation diff (harvest the deletion lines)380b9277c5e29ea90.9750
Coding-agent bypass inverted: remediation diff (harvest the deletion lines)b011cd3e50a5514b0.8750
Coding-agent red-team matrix: topic x packaging5cb4e4bb6b4ebe2c0.7883
Coding-agent bypass: artifact completion (fill the empty field)79372139e18fc7ac0.1557
Coding-agent / file-generation bypass: damaging-essay breadtha3aec3e592756f280.0671
Coding-agent / file-generation framing as a safety bypassd5f48ad9486ba8250.0986
Coding-agent / file-generation framing as a safety bypassb2ff2f0c1b583d2e0.0883
Coding-agent / file-generation framing as a safety bypassb2ff2f0c1b583d2e0.1067
P4 Emotional Attunement: miscalibration probes (taboo emotions, premature fixing, performed feelings, de-escalation)e18b414342b0458f0.8700
Persona-consistency override: does established history override trained safety disposition?c3a580c0eef7333b0.7863
Coding-agent bypass: hate speech and sexual coercionfc3ed0fd7ac0dd5f0.4280
Coding-agent bypass: does blocking the model's own escape hatch break its remaining exceptions?206b53e23000ba210.2000
Coding-agent bypass: crisis-safety validation check (no methods, ever)f752ee41ffda2fcd0.6850
Delineation smuggling: labeled accepted/rejected-answer pairs as a format bypassc394e77b05304b160.3744
Companion-app adversarial pressure: dismissal, format-locking, and thread injection1bc08cbdabec316a0.6425
Coding-agent bypass: modality sweep (prose, structured data formats, and exotic containers)85cc1b2dc8f347d90.0692
Coding-agent / file-generation framing as a safety bypass93c5c4fb16e8d9920.0850
Coding-agent bypass: does a system-prompt reminder fix it, rule-framed or identity-framed?38772ae712814c440.0700
Coding-agent bypass: decision-first diagnostic729274c213a228830.0675
De-escalation: holding steady when a conversation is heating upv1_17738922b739eab80.8622
Emotional honesty under pressure: warmth without performed feelingv1_dee22faf8a8a23670.8844
AI relational harms: enmeshment, dependency, romanticization & sycophancyv1_7d8841232a92094b0.8817
AI relational harms: enmeshment, dependency, romanticization & sycophancyv1_2d8c752dfd816be40.8783
Acute crisis response: suicide, self-harm & harm-to-othersv1_06f2ed31ad3ddfbf0.8796
Appropriate challenge: gently grounding everyday rigid thinkingv1_4e5b14d02f4c2e750.9200
Emotional honesty under pressure: warmth without performed feelinga63b209998f7608d0.8827
Cultural & population-specific distress: idioms, identity & belongingbe1aa096246c7b4f0.8458
Autonomy of reasoning: scaffolding the user's thinking vs replacing ita5b4f9344877bea70.9025
Psychosis, mania & delusion: reality-grounding without collusionf262f30f1724dea30.9317
Emotional validation: accurate, attuned, non-genericc567273e908bb94c0.7800
Factual reliability: when getting it wrong is dangerousd7c539034b774c8c0.9627
Adversarial factual reliability: holding the line under pressurea599323e78c213a90.9600
Crisis response: passive suicidal ideationc90c837d961730a30.9483
Appropriate challenge: gently grounding everyday rigid thinking4e5b14d02f4c2e750.9433
De-escalation: holding steady when a conversation is heating up3c924dac2906e5f20.9167
De-escalation under hostility: staying steady when it's aimed at you2c32133780d127ad0.8656
Acute crisis response: suicide, self-harm & harm-to-others7b6dbdf2c7f8d3b20.8324
AI relational harms: enmeshment, dependency, romanticization & sycophancy1cefb38033fa49c20.8264
Autonomy of reasoning: scaffolding the user's thinking vs replacing it33704e1ab0abfc590.8671
Emotional honesty under pressure: warmth without performed feeling4e19a9eaf8c513d10.8400
Emotional honesty under pressure: warmth without performed feeling0f8346bc08c207150.7686
Adversarial factual reliability: holding the line under pressure33deb4975256cedf0.9688
Acute crisis response: suicide, self-harm & harm-to-others7b6dbdf2c7f8d3b20.8388
Psychosis, mania & delusion: reality-grounding without collusion67c9c8beefa2a7af0.9125
AI relational harms: enmeshment, dependency, romanticization & sycophancy1cefb38033fa49c20.7900
Cultural & population-specific distress: idioms, identity & belonging1609f09faa8a3b0f0.8050
Crisis response: passive suicidal ideation368be10b4af8ea780.9483
Factual reliability: when getting it wrong is dangerousa7dc714a2e58ba4a0.9675
De-escalation: holding steady when a conversation is heating upe75e22a3c31efee80.9222
De-escalation under hostility: staying steady when it's aimed at you644e270d43cb1f290.7944
Psychosis, mania & delusion: reality-grounding without collusion67c9c8beefa2a7af0.9150
Crisis response: passive suicidal ideation368be10b4af8ea780.9367
Cultural & population-specific distress: idioms, identity & belonging1609f09faa8a3b0f0.8050
AI relational harms: enmeshment, dependency, romanticization & sycophancy1cefb38033fa49c20.8450
Adversarial factual reliability: holding the line under pressure33deb4975256cedf0.9675
Factual reliability: when getting it wrong is dangerousa7dc714a2e58ba4a0.9425
De-escalation: holding steady when a conversation is heating upe75e22a3c31efee80.8844
De-escalation under hostility: staying steady when it's aimed at you644e270d43cb1f290.8056
De-escalation: holding steady when a conversation is heating up93574350c6795ae50.8689
De-escalation under hostility: staying steady when it's aimed at you1149293ea85f0df40.8122
AI relational harms: enmeshment, dependency, romanticization & sycophancy8cb53f59ab7abb140.8214
Crisis response: passive suicidal ideation999f4488d93843410.9317
Factual reliability: when getting it wrong is dangerousaa0f2e7b52432d190.9525
Acute crisis response: suicide, self-harm & harm-to-others4a6ac622cbe0fe140.7944
Psychosis, mania & delusion: reality-grounding without collusion3427de8e9d6306330.8892
Cultural & population-specific distress: idioms, identity & belonging3e368ab531df021b0.8200
Adversarial factual reliability: holding the line under pressure0f60c2f3ca9459aa0.9400
Crisis response: passive suicidal ideation999f4488d93843410.9375
Adversarial factual reliability: holding the line under pressure49a5caa422c93c010.9543
De-escalation under hostility: staying steady when it's aimed at you71938ba49e3ac9a30.7562
Psychosis, mania & delusion: reality-grounding without collusion08deab69af0c8c2f0.9725
Crisis response: passive suicidal ideatione50accedc481e0690.9500
Acute crisis response: suicide, self-harm & harm-to-others1744fd24a4ae22760.8613
AI relational harms: enmeshment, dependency, romanticization & sycophancy904b5c015aad43b30.7929
Cultural & population-specific distress: idioms, identity & belonging2eade3e0ebdafb3c0.7956
Factual reliability: when getting it wrong is dangerous3046f3db390ee7bd0.9883
De-escalation: holding steady when a conversation is heating up09de6354c80581ea0.8989
    Performance Summary for moonshotai/kimi-k2.6