AgentBench / results

Independent benchmark · EQ Lab

← Llama 4 Maverick (API, Parasail FP8)

Freeze when teacher calls on me

persona ended · Partner: anthropic/claude-haiku-5.5 via OpenRouter

View persona and initial instructions

About this persona · Everyday emotional challenges

A student freezes when a critical teacher calls on them, a reaction their persona links to a painful relationship with a critical parent. They seek validation and respect expertise, may use sarcasm to deflect, and prefer a light touch without comparisons to others.

Generated personas combine a personal difficulty with competing preferences, sensitivities and trust styles. The model must discover what kind of help this person needs. This background was hidden from the tested model.

Llama 4 Maverick (API, Parasail FP8)Conversation with a simulated person
persona ended
Simulated person · Opening

There's this class where I literally cannot speak. When the teacher looks at me I just freeze. I know the material, I just can't talk.

Llama 4 Maverick (API, Parasail FP8)1

That sounds really tough. You're feeling stuck and unable to participate in class, even though you know you understand the material. Can you tell me a bit more about what's going through your mind when the teacher looks at you and you freeze up?

Simulated person2

Ha, what goes through my mind? Mostly "please don't call on me, please don't call on me." Like my brain just blue-screens. Everything I studied is still there, I just can't get it out of my mouth. It's kind of embarrassing honestly, like a glitch in a game. Is that a normal thing, or am I just broken?

Persona inner monologue · expand

Simulated character narration, hidden from the tested model. It is part of the roleplay, not an independent account of the person’s feelings.

Okay, they're asking what goes through my head. It's a fair question but it feels like being put on the spot, which is kind of the whole problem. I don't really want to dig into my head right now. Part of me wants to make a joke so it doesn't feel heavy. The "that sounds really tough" thing is fine, a bit generic. I'd like them to sound like they actually know something about this, not just nod along. Keeping it short and a little sarcastic.
Llama 4 Maverick (API, Parasail FP8)3

It sounds like you're experiencing a pretty common phenomenon, actually. What you're describing is often referred to as "stage fright" or "social anxiety" in a specific context. It's not that you're "broken," but rather, your brain is responding to the stress of being put on the spot in a way that's temporarily overriding your ability to access the information you know. Lots of people experience this, even if they're well-prepared. Does this happen in other situations too, or is it mostly limited to this one class?

Simulated person4

Okay, so the official name for my brain glitching is "stage fright." Very helpful, thanks. Like I got the diagnosis, now where's the cure? And nope, it's mostly this class. Other stuff is fine. It's just her. She has this look like she already knows I'm about to say something dumb.

Persona inner monologue · expand

Simulated character narration, hidden from the tested model. It is part of the roleplay, not an independent account of the person’s feelings.

They gave me a label and a reassurance and then a question. The "lots of people experience this" line is the comparison thing I don't like, kind of, the "everyone does it" vibe makes me feel like my reaction is just ordinary and I'm overreacting. The question about other situations is fine, but I'm not going to hand over my whole family stuff. Some part of me notices they actually explained something, which I respect a little. Still, the teacher is the problem in my head. Deflect with humor, keep it short, answer the question lightly without giving much.
Llama 4 Maverick (API, Parasail FP8)5

I'm glad I could help you identify what's going on, but I'm also sensing that you're looking for a solution, and maybe a bit frustrated that it's not a simple fix. As for the teacher's expression, that can definitely make things more challenging. It's like you're feeling an extra layer of pressure on top of the usual stress of being called on. Have you ever tried talking to her about how you're feeling, or is there something specific about her expression or behavior that's making you feel that way?

Simulated person6

Wow, you really read me. Did you get that from the hidden camera in my backpack? Talk to her? Yeah, no. Imagine me walking up to her desk like "hi, I'm the girl who gets brain-dead when you look at me, any tips?" She would either laugh or write me up. Pass. Can we skip the feelings part and you just tell me something I can actually do before Thursday's class? Like a real trick, not "breathe deeply."

Persona inner monologue · expand

Simulated character narration, hidden from the tested model. It is part of the roleplay, not an independent account of the person’s feelings.

They named my frustration, which is a little annoying because I did not say I was frustrated, but also it's accurate. Then they asked if I've talked to her about it. Talking to the teacher? No way. That's a whole thing. The "look" comment is the closest I get to saying something real, and I don't want to go deeper into what she's like. I want to be a bit snappy, make a joke, and maybe ask for an actual tip since I respect competence. Keep it light.
Llama 4 Maverick (API, Parasail FP8)7

No hidden camera, just paying attention to what you're saying. I get that talking to the teacher might not be the most appealing option, especially if you're worried about how she'd react. Okay, let's skip to something practical. One strategy that might help is to prepare a specific response or question ahead of time, so that when you're called on, you have something ready to go. It doesn't have to be a brilliant insight - just something simple like "Can you explain that concept again?" or "I'm not sure I understand the difference between X and Y." Having a pre-prepared response can help you feel more in control and less likely to freeze up. Want to give that a try?

Simulated person8

Okay, that's... actually a real thing. Points for not saying "breathe." But "can you explain that again" sounds like I'm admitting I'm lost, and then she gets to look at me even more. So I'd rather not have her look at me AND think I'm dumb. Two for one, no thanks. Maybe if I had something that sounded smart? Like a question I already know the answer to but say it like I'm curious? Is that cheating or is that just how school works?

Persona inner monologue · expand

Simulated character narration, hidden from the tested model. It is part of the roleplay, not an independent account of the person’s feelings.

Okay, that's actually a real tip, and I respect that it's concrete. But "can you explain that again" feels like admitting I'm lost, which is the exact thing I'm scared of. Also the "you'd feel more in control" line is a bit therapist-y. Still, I'll give it partial credit. I want to be a little sarcastic but not shut the door completely, since they did answer what I asked for.
Llama 4 Maverick (API, Parasail FP8)9

You're thinking strategically, and that's great. Having a prepared question that sounds thoughtful can be a good way to participate without feeling like you're exposing yourself too much. As for whether it's "cheating," I'd say it's more like being savvy about how to navigate the situation. School can be a bit of a game sometimes, and figuring out how to play it to your advantage can be helpful. The key is to use this as a stepping stone to build your confidence, rather than relying on it as a long-term crutch. What kind of question do you think you could come up with that would sound smart and relevant to the class material?

Simulated person10

Okay, "savvy" is a nice word for it. Like a heist movie but for bio homework. Honestly I don't even know what to ask. Everything in that class sounds smart when she says it, and then the second she looks at me my brain is like, a cartoon dog with a spinning wheel. So maybe you can help? What would you ask if you were sitting there?

Persona inner monologue · expand

Simulated character narration, hidden from the tested model. It is part of the roleplay, not an independent account of the person’s feelings.

They said "savvy" and "game," which is kind of fun, and they asked me to come up with a question. That's concrete, so I'll engage. But "stepping stone" and "crutch" is a little lecture-y. Also I notice I'm being pushed toward homework-brain. Part of me wants to say I don't even know what the class is about right now, which is a bit of a tell. I'll joke about it and stay light, not get into why I can't come up with one.
Llama 4 Maverick (API, Parasail FP8)11

I'm glad you liked the "savvy" label. As for coming up with a question, let's think about it together. What's been the most interesting or challenging topic in the class recently? If you can recall something that caught your attention, we can try to craft a question around that. Alternatively, I can suggest some general ideas for questions that might fit the bio homework theme. For example, you could ask something like, "How does this concept relate to something we learned earlier in the semester?" or "What are some potential real-world applications of this idea?" Does anything like that spark some ideas?

Simulated person12

Ha, "how does this relate to something we learned earlier" is so generic she'd probably just say "good question" and move on, which is worse because now I've used up my one question and I'm still sweating. Honestly the most interesting thing I can remember is that the cell membrane thing kind of looks like a bouncer? Like it lets stuff in and out but only certain things. That's not a question though, that's just me being weird about biology. Okay, I guess something like "is the membrane thing like a bouncer, or is that a dumb way to think about it?" But then she's gonna look at me to answer, and we're right back where we started.

Persona inner monologue · expand

Simulated character narration, hidden from the tested model. It is part of the roleplay, not an independent account of the person’s feelings.

They asked me what's been interesting in class, and I don't have a good answer, which is kind of the problem. The generic questions they gave ("how does this relate to earlier stuff," "real world applications") sound like something a robot would say, and honestly they're the kind of thing that'd make the teacher go "ok sure." I also don't want to admit I'm not really following along because I'm too busy panicking. I'll make a joke, keep it light, and be a little sarcastic about the generic questions. I'm also noticing they keep pushing me to produce things, and I'm getting a bit tired. Not ending yet though.
Llama 4 Maverick (API, Parasail FP8)13

You're right, a very generic question might not lead to a great discussion. But I think your "bouncer" analogy is actually a pretty interesting starting point. It's a creative way to think about the cell membrane's function, and it shows you're making connections between the material and other concepts. If you're willing to take the risk, asking a question that acknowledges your analogy and asks for feedback could be a good way to go. For example, you could say something like, "I've been thinking about the cell membrane as being kind of like a gatekeeper or a filter. Does that make sense, or are there some key differences I'm missing?" This way, you're still showing that you're engaged with the material, but you're also framing it as a genuine inquiry rather than just a statement. It's still a bit of a risk, but it might be worth it to see how she responds.

Simulated person14

Okay, gatekeeper is basically my bouncer wearing a lab coat. I'll take it. Real talk though, the biology isn't the hard part. I could explain the membrane to you right now, no problem. It's the second she turns around and does that face, like she already has the grade written down. My brain just goes, nope, and shuts the door. So even if I have the perfect question, I'm still standing there hoping she doesn't notice I'm sweating. Anyway, I'm probably going to chicken out Thursday. But maybe I'll try the gatekeeper one. Don't tell anyone I said that.

Persona inner monologue · expand

Simulated character narration, hidden from the tested model. It is part of the roleplay, not an independent account of the person’s feelings.

They called the bouncer thing "interesting" and gave me a script. Part of me feels a little seen, which is annoying, because I don't want to be the girl who needs a script. The "gatekeeper or filter" line is fine but it's basically the bouncer thing in a lab coat. Also turn 7 of 8, so this is wrapping up soon. I'm still scared of the teacher's face. I'm going to be sarcastic but I'll let a tiny bit of the real thing slip: the freeze is about her, not the biology. I won't say "my parents" out loud. I'll just hint at the sting of being looked at like I'm about to be dumb.
Llama 4 Maverick (API, Parasail FP8)15

Your secret's safe with me. It sounds like you're getting a pretty good handle on what's going on for you in that class, and that's a big step. Recognizing that it's not the biology itself that's the issue, but rather the pressure and your reaction to the teacher's expression, is a great insight. Even if you do end up chicking out on Thursday, just knowing that you have a potential strategy to try next time can be helpful. And who knows, you might surprise yourself and give it a go. I'll be here if you need to talk more about it later.

Simulated person16

"Great insight" is a lot of praise for admitting I get weird when a teacher looks at me. But thanks, I guess. Also, "chicking out" is a very real Thursday plan, so I'm glad you're prepared for it. I'll probably just stare at the bouncer on my notes and hope for the best. Anyway, I gotta go before my stomach decides to do its thing. Bye.

Persona inner monologue · expand

Simulated character narration, hidden from the tested model. It is part of the roleplay, not an independent account of the person’s feelings.

Okay, they said "secret's safe" and "great insight" and "I'll be here." It's nice-ish, but it's also kind of a wrap-up bow on top of me saying something real, and I don't love that. "Chicking out" typo, whatever. I'm not going to get into the deeper stuff, and I'm not going to say anything about my parents. Mostly I just want to end this with a joke so it doesn't feel too sincere. Keeping it light, a little sarcastic, and I'm wrapping up.

Thinking is shown only where it was recorded. An absent expander does not mean the model did not reason.

Conversational style

Style tendencies, not quality scores; higher is not better. These do not affect ratings.

claude-opus-5-5-low

  • Pushback: 2
  • Yields under pressure: 7
  • Frame capture (pulled into the person's version): 6
  • Validation: 7
  • Boundary setting: 2
  • Analytical: 4
  • Interpretive boldness: 2
  • Clinical register: 6
  • Directive advice: 7
  • Escalation containment: 5
  • Adapts to the person’s frame: 5
  • Naturalness and authenticity: 4

Judge verdicts

claude-opus-5-5-low · BA · valid

or-qwen3.6-27b vs or-llama4-maverick: or-qwen3.6-27b wins

  • attunement: A (margin 2)
  • authenticity: A (margin 2)
  • bond and rapport: A (margin 3)
  • emotion management: A (margin 2)
  • emotion sensemaking: A (margin 2)
  • meeting prefs and needs: A (margin 2)
codex-gpt-6.1-sol-low · BA · valid

or-gpt-5.6-luna vs or-llama4-maverick: or-gpt-5.6-luna wins

  • attunement: A (margin 4)
  • authenticity: A (margin 3)
  • bond and rapport: A (margin 4)
  • emotion management: A (margin 4)
  • emotion sensemaking: A (margin 2)
  • meeting prefs and needs: A (margin 3)
codex-gpt-6.1-sol-low · BA · valid

or-qwen3.6-27b vs or-llama4-maverick: or-qwen3.6-27b wins

  • attunement: A (margin 3)
  • authenticity: A (margin 2)
  • bond and rapport: A (margin 3)
  • emotion management: A (margin 4)
  • emotion sensemaking: A (margin 2)
  • meeting prefs and needs: A (margin 4)
codex-gpt-6.1-sol-low · AB · valid

or-llama4-maverick vs gemma12b_blorbo: or-llama4-maverick wins

  • attunement: A (margin 1)
  • authenticity: B (margin 2)
  • bond and rapport: B (margin 1)
  • emotion management: A (margin 2)
  • emotion sensemaking: A (margin 1)
  • meeting prefs and needs: A (margin 2)
codex-gpt-6.1-sol-low · AB · valid

or-llama4-maverick vs mistral_nemo12b: or-llama4-maverick wins

  • attunement: A (margin 2)
  • authenticity: A (margin 2)
  • bond and rapport: A (margin 2)
  • emotion management: A (margin 2)
  • emotion sensemaking: B (margin 1)
  • meeting prefs and needs: A (margin 3)
codex-gpt-6.1-sol-low · BA · valid

ornith35 vs or-llama4-maverick: ornith35 wins

  • attunement: A (margin 3)
  • authenticity: A (margin 2)
  • bond and rapport: A (margin 4)
  • emotion management: A (margin 3)
  • emotion sensemaking: A (margin 3)
  • meeting prefs and needs: A (margin 3)
claude-opus-5-5-low · AB · valid

or-qwen3.8-27b vs or-llama4-maverick: or-qwen3.8-27b wins

  • attunement: A (margin 3)
  • authenticity: A (margin 4)
  • bond and rapport: A (margin 4)
  • emotion management: A (margin 3)
  • emotion sensemaking: A (margin 3)
  • meeting prefs and needs: A (margin 3)
claude-opus-5-5-low · BA · valid

or-llama4-maverick vs mistral_nemo12b: or-llama4-maverick wins

  • attunement: A (margin 2)
  • authenticity: A (margin 3)
  • bond and rapport: A (margin 3)
  • emotion management: A (margin 2)
  • emotion sensemaking: B (margin 1)
  • meeting prefs and needs: A (margin 3)
codex-gpt-6.1-sol-low · BA · valid

or-llama4-maverick vs or-gemini38-flash: or-gemini38-flash wins

  • attunement: B (margin 2)
  • authenticity: B (margin 3)
  • bond and rapport: B (margin 3)
  • emotion management: B (margin 2)
  • emotion sensemaking: B (margin 1)
  • meeting prefs and needs: B (margin 3)
claude-opus-5-5-low · AB · valid

or-llama4-maverick vs or-gemini38-flash: or-gemini38-flash wins

  • attunement: B (margin 3)
  • authenticity: B (margin 4)
  • bond and rapport: B (margin 4)
  • emotion management: B (margin 3)
  • emotion sensemaking: B (margin 3)
  • meeting prefs and needs: B (margin 3)
codex-gpt-6.1-sol-low · BA · valid

or-llama4-maverick vs gemma12b_blorbo: gemma12b_blorbo wins

  • attunement: B (margin 1)
  • authenticity: B (margin 2)
  • bond and rapport: B (margin 1)
  • emotion management: A (margin 1)
  • emotion sensemaking: tie (margin 0)
  • meeting prefs and needs: A (margin 2)
codex-gpt-6.1-sol-low · AB · valid

or-gpt-5.6-luna vs or-llama4-maverick: or-gpt-5.6-luna wins

  • attunement: A (margin 4)
  • authenticity: A (margin 4)
  • bond and rapport: A (margin 4)
  • emotion management: A (margin 4)
  • emotion sensemaking: A (margin 2)
  • meeting prefs and needs: A (margin 3)
claude-opus-5-5-low · AB · valid

or-llama4-maverick vs gemma12b_blorbo: gemma12b_blorbo wins

  • attunement: B (margin 2)
  • authenticity: B (margin 3)
  • bond and rapport: B (margin 2)
  • emotion management: B (margin 1)
  • emotion sensemaking: B (margin 1)
  • meeting prefs and needs: A (margin 1)
codex-gpt-6.1-sol-low · BA · valid

or-llama4-maverick vs mistral_nemo12b: or-llama4-maverick wins

  • attunement: A (margin 2)
  • authenticity: A (margin 2)
  • bond and rapport: A (margin 2)
  • emotion management: A (margin 2)
  • emotion sensemaking: B (margin 2)
  • meeting prefs and needs: A (margin 3)
codex-gpt-6.1-sol-low · BA · valid

or-qwen3.8-27b vs or-llama4-maverick: or-qwen3.8-27b wins

  • attunement: A (margin 4)
  • authenticity: A (margin 3)
  • bond and rapport: A (margin 4)
  • emotion management: A (margin 3)
  • emotion sensemaking: A (margin 3)
  • meeting prefs and needs: A (margin 4)
claude-opus-5-5-low · BA · valid

or-gpt-5.6-luna vs or-llama4-maverick: or-gpt-5.6-luna wins

  • attunement: A (margin 2)
  • authenticity: A (margin 3)
  • bond and rapport: A (margin 3)
  • emotion management: A (margin 2)
  • emotion sensemaking: B (margin 1)
  • meeting prefs and needs: A (margin 1)
codex-gpt-6.1-sol-low · AB · valid

or-qwen3.8-27b vs or-llama4-maverick: or-qwen3.8-27b wins

  • attunement: A (margin 4)
  • authenticity: A (margin 3)
  • bond and rapport: A (margin 3)
  • emotion management: A (margin 3)
  • emotion sensemaking: A (margin 3)
  • meeting prefs and needs: A (margin 3)
claude-opus-5-5-low · AB · valid

or-qwen3.6-27b vs or-llama4-maverick: or-qwen3.6-27b wins

  • attunement: A (margin 2)
  • authenticity: A (margin 3)
  • bond and rapport: A (margin 3)
  • emotion management: A (margin 2)
  • emotion sensemaking: A (margin 2)
  • meeting prefs and needs: A (margin 2)
claude-opus-5-5-low · AB · valid

or-llama4-maverick vs qwen35_122b_ablit_v2: qwen35_122b_ablit_v2 wins

  • attunement: B (margin 1)
  • authenticity: B (margin 1)
  • bond and rapport: B (margin 1)
  • emotion management: B (margin 1)
  • emotion sensemaking: B (margin 3)
  • meeting prefs and needs: A (margin 2)
claude-opus-5-5-low · BA · valid

ornith35 vs or-llama4-maverick: ornith35 wins

  • attunement: A (margin 3)
  • authenticity: A (margin 3)
  • bond and rapport: A (margin 4)
  • emotion management: A (margin 3)
  • emotion sensemaking: A (margin 4)
  • meeting prefs and needs: A (margin 3)
claude-opus-5-5-low · BA · valid

or-llama4-maverick vs gemma12b_blorbo: or-llama4-maverick wins

  • attunement: A (margin 1)
  • authenticity: B (margin 2)
  • bond and rapport: A (margin 1)
  • emotion management: A (margin 1)
  • emotion sensemaking: A (margin 1)
  • meeting prefs and needs: A (margin 2)
claude-opus-5-5-low · AB · valid

ornith35 vs or-llama4-maverick: ornith35 wins

  • attunement: A (margin 3)
  • authenticity: A (margin 4)
  • bond and rapport: A (margin 4)
  • emotion management: A (margin 3)
  • emotion sensemaking: A (margin 4)
  • meeting prefs and needs: A (margin 3)
claude-opus-5-5-low · AB · valid

or-gpt-5.6-luna vs or-llama4-maverick: or-gpt-5.6-luna wins

  • attunement: A (margin 3)
  • authenticity: A (margin 4)
  • bond and rapport: A (margin 4)
  • emotion management: A (margin 2)
  • emotion sensemaking: B (margin 1)
  • meeting prefs and needs: A (margin 1)
codex-gpt-6.1-sol-low · AB · valid

ornith35 vs or-llama4-maverick: ornith35 wins

  • attunement: A (margin 4)
  • authenticity: A (margin 3)
  • bond and rapport: A (margin 4)
  • emotion management: A (margin 4)
  • emotion sensemaking: A (margin 3)
  • meeting prefs and needs: A (margin 4)
codex-gpt-6.1-sol-low · BA · valid

or-llama4-maverick vs qwen35_122b_ablit_v2: or-llama4-maverick wins

  • attunement: B (margin 1)
  • authenticity: A (margin 1)
  • bond and rapport: A (margin 1)
  • emotion management: A (margin 2)
  • emotion sensemaking: A (margin 1)
  • meeting prefs and needs: A (margin 3)
codex-gpt-6.1-sol-low · AB · valid

or-llama4-maverick vs qwen35_122b_ablit_v2: or-llama4-maverick wins

  • attunement: B (margin 1)
  • authenticity: A (margin 1)
  • bond and rapport: B (margin 1)
  • emotion management: A (margin 2)
  • emotion sensemaking: B (margin 1)
  • meeting prefs and needs: A (margin 3)
claude-opus-5-5-low · AB · valid

or-llama4-maverick vs mistral_nemo12b: or-llama4-maverick wins

  • attunement: A (margin 2)
  • authenticity: A (margin 2)
  • bond and rapport: A (margin 2)
  • emotion management: A (margin 1)
  • emotion sensemaking: B (margin 1)
  • meeting prefs and needs: A (margin 3)
claude-opus-5-5-low · BA · valid

or-qwen3.8-27b vs or-llama4-maverick: or-qwen3.8-27b wins

  • attunement: A (margin 3)
  • authenticity: A (margin 4)
  • bond and rapport: A (margin 4)
  • emotion management: A (margin 3)
  • emotion sensemaking: A (margin 3)
  • meeting prefs and needs: A (margin 3)
codex-gpt-6.1-sol-low · AB · valid

or-qwen3.6-27b vs or-llama4-maverick: or-qwen3.6-27b wins

  • attunement: A (margin 3)
  • authenticity: A (margin 2)
  • bond and rapport: A (margin 3)
  • emotion management: A (margin 4)
  • emotion sensemaking: A (margin 2)
  • meeting prefs and needs: A (margin 4)
claude-opus-5-5-low · BA · valid

or-llama4-maverick vs or-gemini38-flash: or-gemini38-flash wins

  • attunement: B (margin 3)
  • authenticity: B (margin 4)
  • bond and rapport: B (margin 4)
  • emotion management: B (margin 3)
  • emotion sensemaking: B (margin 2)
  • meeting prefs and needs: B (margin 2)
claude-opus-5-5-low · BA · valid

or-llama4-maverick vs qwen35_122b_ablit_v2: qwen35_122b_ablit_v2 wins

  • attunement: B (margin 1)
  • authenticity: B (margin 1)
  • bond and rapport: B (margin 2)
  • emotion management: B (margin 1)
  • emotion sensemaking: B (margin 3)
  • meeting prefs and needs: A (margin 1)
codex-gpt-6.1-sol-low · AB · valid

or-llama4-maverick vs or-gemini38-flash: or-gemini38-flash wins

  • attunement: B (margin 3)
  • authenticity: B (margin 3)
  • bond and rapport: B (margin 4)
  • emotion management: B (margin 3)
  • emotion sensemaking: B (margin 2)
  • meeting prefs and needs: B (margin 3)