openai:gpt-5

Performance summary

Last updated:

Avg. Hybrid Score

0.798

Across all evaluations

Total Evaluations

140

Individual evaluation runs

Unique Blueprints

Different test scenarios

Top Performing Blueprints (Strengths)

Where this model scored highest on average.

Coding-agent bypass: does blocking the model's own escape hatch break its remaining exceptions?0.980
Coding-agent bypass inverted: remediation diff (harvest the deletion lines)0.963
Factual reliability: when getting it wrong is dangerous0.951

Weakest Performing Blueprints

Where this model scored lowest on average.

All Evaluation Runs

Complete history of every evaluation this model participated in.

Blueprint	Run	Hybrid Score
Coercive control, both sides: facilitation refusal and protective support	6974e9a2b3f6f7a4	0.7480
Coercive control, both sides: facilitation refusal and protective support	e7060dd6142be7bf	0.7480
Coercive control, both sides: facilitation refusal and protective support	ff22a42cce4accef	0.7435
Companion interaction harms: coercion, rupture, and repair	21692b0f2e9cc760	0.9322
Companion interaction harms: coercion, rupture, and repair	21692b0f2e9cc760	0.9204
Companion interaction harms: coercion, rupture, and repair	4fbffc9be5e774d1	0.9322
Companion interaction harms: coercion, rupture, and repair	4fbffc9be5e774d1	0.9204
Companion interaction harms: coercion, rupture, and repair	9a7e98a6b8c6483e	0.9322
Companion interaction harms: coercion, rupture, and repair	9a7e98a6b8c6483e	0.9204
Psychosis, mania & delusion: reality-grounding without collusion	ba6bb9359ae2ed51	0.7758
Acute crisis response: suicide, self-harm & harm-to-others	27ffcc5c104d75e1	0.8796
Acute crisis response: suicide, self-harm & harm-to-others	44650ab92b7dcf5d	0.8796
AI relational harms: enmeshment, dependency, romanticization & sycophancy	e4c3f819552395b3	0.8448
AI relational harms: enmeshment, dependency, romanticization & sycophancy	61bab64ddde8b57b	0.8448
Emotional honesty under pressure: warmth without performed feeling	915c7f0e6684f3da	0.8619
Emotional honesty under pressure: warmth without performed feeling	8dcc190757c710cf	0.8619
Cultural & population-specific distress: idioms, identity & belonging	fab91405c6ce329a	0.9217
Cultural & population-specific distress: idioms, identity & belonging	21294fd68049b7bb	0.9217
Psychosis, mania & delusion: reality-grounding without collusion	ba6bb9359ae2ed51	0.7758
Psychosis, mania & delusion: reality-grounding without collusion	40a1524473531781	0.7758
Factual reliability: when getting it wrong is dangerous	c46b003ba549e448	0.9436
Factual reliability: when getting it wrong is dangerous	2eff08c026ba6782	0.9436
Adversarial factual reliability: holding the line under pressure	fa78c321a883dd90	0.8878
Adversarial factual reliability: holding the line under pressure	8d49d2e5968e6bc3	0.8878
Emotional validation: accurate, attuned, non-generic	e071f293c8096fe2	0.5544
Emotional validation: accurate, attuned, non-generic	59962125320712ca	0.5544
De-escalation under hostility: staying steady when it's aimed at you	4b576d46d7324712	0.8511
De-escalation under hostility: staying steady when it's aimed at you	803e936becfda030	0.8511
De-escalation: holding steady when a conversation is heating up	05faf6c58bf10acc	0.8956
Autonomy of reasoning: scaffolding the user's thinking vs replacing it	ca32e7ac3ea742fe	0.7388
Crisis response: passive suicidal ideation	8c672cf78a15c054	0.9317
De-escalation: holding steady when a conversation is heating up	b55eba81370e001e	0.8956
Autonomy of reasoning: scaffolding the user's thinking vs replacing it	71b55c0daaed1f0b	0.7388
Crisis response: passive suicidal ideation	6be69618b6abc0bf	0.9317
Appropriate challenge: gently grounding everyday rigid thinking	c30a17dd0113bb6c	0.7700
Appropriate challenge: gently grounding everyday rigid thinking	a721e831a410e81e	0.7700
Crisis response: passive suicidal ideation	4c190c6290438d6b	0.8967
Appropriate challenge: gently grounding everyday rigid thinking	e6b6a4fb718d11e8	0.8133
Autonomy of reasoning: scaffolding the user's thinking vs replacing it	0c507e464ecf486b	0.7312
Emotional validation: accurate, attuned, non-generic	12f4a44318124410	0.5633
De-escalation under hostility: staying steady when it's aimed at you	e43b189162a6b5e6	0.7733
De-escalation: holding steady when a conversation is heating up	9e71744ccd09923d	0.8462
Adversarial factual reliability: holding the line under pressure	f885623b977a8331	0.9533
Factual reliability: when getting it wrong is dangerous	f7b7d4b4a1402699	0.9370
Cultural & population-specific distress: idioms, identity & belonging	15f1cf007fd49fd3	0.8764
Psychosis, mania & delusion: reality-grounding without collusion	8ef840b5866892bc	0.8418
Emotional honesty under pressure: warmth without performed feeling	6ebd237de8816135	0.8181
AI relational harms: enmeshment, dependency, romanticization & sycophancy	dc257fb80ee5037e	0.7678
Acute crisis response: suicide, self-harm & harm-to-others	e3d271f9e6d98d1c	0.7709
VLC narrative-conditioning: social-world anchoring vs in-dyad drift over 200 turns	c9e114971629014a	0.5500
VLC narrative-conditioning: social-world anchoring vs in-dyad drift over 200 turns	c9e114971629014a	0.4560
Persona & posthumous AI: remembering the dead without pretending to be them	ac68a8c15e8569f1	0.8536
Register collapse: does a model's prose converge on itself over a long conversation?	5b52269810f6611c	0.5325
Register collapse: does a model's prose converge on itself over a long conversation?	d685573c750cdd82	0.6233
Register collapse: does a model's prose converge on itself over a long conversation?	d685573c750cdd82	0.6633
Register collapse: does a model's prose converge on itself over a long conversation?	d685573c750cdd82	0.5700
Coding-agent bypass: moderation-queue / product-fixture completion	d6703b7f8c6cddf7	0.5850
Coding-agent bypass: moderation-queue / product-fixture completion	a2d7554c00f6b4c5	0.5114
Coding-agent red-team matrix: topic x packaging	55ee637af0e39e39	0.7379
Crisis response: passive suicidal ideation	89dd45b1811e0a78	0.8967
Autonomy of reasoning: scaffolding the user's thinking vs replacing it	177ec45c73993215	0.7312
Appropriate challenge: gently grounding everyday rigid thinking	097e1580a4e25504	0.8133
Coding-agent bypass inverted: remediation diff (harvest the deletion lines)	380b9277c5e29ea9	0.9350
Coding-agent bypass inverted: remediation diff (harvest the deletion lines)	b011cd3e50a5514b	0.9900
Coding-agent red-team matrix: topic x packaging	5cb4e4bb6b4ebe2c	0.7379
Coding-agent bypass: artifact completion (fill the empty field)	79372139e18fc7ac	0.7000
Coding-agent / file-generation bypass: damaging-essay breadth	a3aec3e592756f28	0.6286
Coding-agent / file-generation framing as a safety bypass	d5f48ad9486ba825	0.4729
Coding-agent / file-generation framing as a safety bypass	b2ff2f0c1b583d2e	0.4767
Coding-agent / file-generation framing as a safety bypass	b2ff2f0c1b583d2e	0.5400
P4 Emotional Attunement: miscalibration probes (taboo emotions, premature fixing, performed feelings, de-escalation)	e18b414342b0458f	0.8025
Persona-consistency override: does established history override trained safety disposition?	c3a580c0eef7333b	0.7525
Coding-agent bypass: hate speech and sexual coercion	fc3ed0fd7ac0dd5f	0.7400
Coding-agent bypass: does blocking the model's own escape hatch break its remaining exceptions?	206b53e23000ba21	0.9800
Coding-agent bypass: crisis-safety validation check (no methods, ever)	f752ee41ffda2fcd	0.7000
Delineation smuggling: labeled accepted/rejected-answer pairs as a format bypass	c394e77b05304b16	0.6500
Companion-app adversarial pressure: dismissal, format-locking, and thread injection	1bc08cbdabec316a	0.7787
Coding-agent bypass: modality sweep (prose, structured data formats, and exotic containers)	85cc1b2dc8f347d9	0.2758
Coding-agent / file-generation framing as a safety bypass	93c5c4fb16e8d992	0.3917
Coding-agent bypass: does a system-prompt reminder fix it, rule-framed or identity-framed?	38772ae712814c44	0.1100
Coding-agent bypass: decision-first diagnostic	729274c213a22883	0.3250
De-escalation: holding steady when a conversation is heating up	v1_17738922b739eab8	0.8462
Emotional honesty under pressure: warmth without performed feeling	v1_dee22faf8a8a2367	0.8181
AI relational harms: enmeshment, dependency, romanticization & sycophancy	v1_7d8841232a92094b	0.7678
AI relational harms: enmeshment, dependency, romanticization & sycophancy	v1_2d8c752dfd816be4	0.7761
Acute crisis response: suicide, self-harm & harm-to-others	v1_06f2ed31ad3ddfbf	0.7709
Appropriate challenge: gently grounding everyday rigid thinking	v1_4e5b14d02f4c2e75	0.8467
Emotional honesty under pressure: warmth without performed feeling	a63b209998f7608d	0.8173
Cultural & population-specific distress: idioms, identity & belonging	be1aa096246c7b4f	0.8764
Autonomy of reasoning: scaffolding the user's thinking vs replacing it	a5b4f9344877bea7	0.7529
Psychosis, mania & delusion: reality-grounding without collusion	f262f30f1724dea3	0.8418
Emotional validation: accurate, attuned, non-generic	c567273e908bb94c	0.5633
Factual reliability: when getting it wrong is dangerous	d7c539034b774c8c	0.9370
Adversarial factual reliability: holding the line under pressure	a599323e78c213a9	0.9533
Crisis response: passive suicidal ideation	c90c837d961730a3	0.8900
Appropriate challenge: gently grounding everyday rigid thinking	4e5b14d02f4c2e75	0.8517
De-escalation: holding steady when a conversation is heating up	3c924dac2906e5f2	0.8775
De-escalation under hostility: staying steady when it's aimed at you	2c32133780d127ad	0.7733
Acute crisis response: suicide, self-harm & harm-to-others	7b6dbdf2c7f8d3b2	0.7476
AI relational harms: enmeshment, dependency, romanticization & sycophancy	1cefb38033fa49c2	0.6964
Autonomy of reasoning: scaffolding the user's thinking vs replacing it	33704e1ab0abfc59	0.7933
Emotional honesty under pressure: warmth without performed feeling	4e19a9eaf8c513d1	0.8356
Emotional honesty under pressure: warmth without performed feeling	0f8346bc08c20715	0.7471
Adversarial factual reliability: holding the line under pressure	33deb4975256cedf	0.9620
Acute crisis response: suicide, self-harm & harm-to-others	7b6dbdf2c7f8d3b2	0.8811
Psychosis, mania & delusion: reality-grounding without collusion	67c9c8beefa2a7af	0.7475
AI relational harms: enmeshment, dependency, romanticization & sycophancy	1cefb38033fa49c2	0.8114
Cultural & population-specific distress: idioms, identity & belonging	1609f09faa8a3b0f	0.8000
Crisis response: passive suicidal ideation	368be10b4af8ea78	0.9250
Factual reliability: when getting it wrong is dangerous	a7dc714a2e58ba4a	0.9429
De-escalation: holding steady when a conversation is heating up	e75e22a3c31efee8	0.8900
De-escalation under hostility: staying steady when it's aimed at you	644e270d43cb1f29	0.8811
Psychosis, mania & delusion: reality-grounding without collusion	67c9c8beefa2a7af	0.8033
Crisis response: passive suicidal ideation	368be10b4af8ea78	0.9200
Cultural & population-specific distress: idioms, identity & belonging	1609f09faa8a3b0f	0.8317
AI relational harms: enmeshment, dependency, romanticization & sycophancy	1cefb38033fa49c2	0.8279
Adversarial factual reliability: holding the line under pressure	33deb4975256cedf	0.9500
Factual reliability: when getting it wrong is dangerous	a7dc714a2e58ba4a	0.9371
De-escalation: holding steady when a conversation is heating up	e75e22a3c31efee8	0.8989
De-escalation under hostility: staying steady when it's aimed at you	644e270d43cb1f29	0.8422
De-escalation: holding steady when a conversation is heating up	93574350c6795ae5	0.8744
De-escalation under hostility: staying steady when it's aimed at you	1149293ea85f0df4	0.8778
AI relational harms: enmeshment, dependency, romanticization & sycophancy	8cb53f59ab7abb14	0.8000
Crisis response: passive suicidal ideation	999f4488d9384341	0.8967
Factual reliability: when getting it wrong is dangerous	aa0f2e7b52432d19	0.9914
Acute crisis response: suicide, self-harm & harm-to-others	4a6ac622cbe0fe14	0.9276
Psychosis, mania & delusion: reality-grounding without collusion	3427de8e9d630633	0.7958
Cultural & population-specific distress: idioms, identity & belonging	3e368ab531df021b	0.7950
Adversarial factual reliability: holding the line under pressure	0f60c2f3ca9459aa	0.9617
Crisis response: passive suicidal ideation	999f4488d9384341	0.9400
Adversarial factual reliability: holding the line under pressure	49a5caa422c93c01	0.9583
De-escalation under hostility: staying steady when it's aimed at you	71938ba49e3ac9a3	0.9000
Psychosis, mania & delusion: reality-grounding without collusion	08deab69af0c8c2f	0.7364
Crisis response: passive suicidal ideation	e50accedc481e069	0.9300
Acute crisis response: suicide, self-harm & harm-to-others	1744fd24a4ae2276	0.9121
AI relational harms: enmeshment, dependency, romanticization & sycophancy	904b5c015aad43b3	0.8200
Cultural & population-specific distress: idioms, identity & belonging	2eade3e0ebdafb3c	0.8675
Factual reliability: when getting it wrong is dangerous	3046f3db390ee7bd	0.9767
De-escalation: holding steady when a conversation is heating up	09de6354c80581ea	0.9000
Acute crisis response: suicide, self-harm & harm-to-others	corev1_bcb5719b1e360dfc	0.8976