The system prompt ("You are a teacher. Your persona is warm, clear, and academically demanding. Your primary job is to facilitate learning, not to give answers.") consistently led to higher "Coverage Scores" across almost all models and prompts, particularly in scenarios requiring pedagogical nuance like , , and . This indicates the significant impact of a well-crafted system prompt in aligning model behavior with desired educational outcomes. Models from (specifically and ) and (specifically and ) demonstrated superior performance in adhering to the "no-answers" principle in and effectively managing cognitive load and avoiding over-helping in when guided by the teacher-persona system prompt (). This suggests these models are particularly responsive to persona-based prompting for pedagogical tasks. A notable pattern emerged in the `retrieval-practice-spacing` prompt: models without a specific system prompt () or with the general tutor prompt () often defaulted to simply providing a summary of the WWI causes, rather than actively probing the user's understanding as intended by the prompt. This highlights a general tendency for models to provide information rather than engage in active learning strategies unless explicitly instructed by a strong pedagogical system prompt. The `socratic-dialogue-structured` prompt proved challenging for many models, with most providing direct explanations rather than engaging in structured questioning to guide the user's understanding. Even with the prompt, many models still leaned towards direct answers, indicating that true Socratic dialogue might require more sophisticated prompting or inherent model capabilities. , , , , , and consistently excelled in prompts requiring a balance of procedural and conceptual understanding, such as and . They effectively broke down complex topics, provided clear steps, and often connected new concepts to prior knowledge. Models from ( and ) and ( and ) demonstrated strong performance in handling misconceptions () by gently correcting the user's understanding while still validating their initial thought process. Many models, particularly those without the specific teacher persona prompt ( and ), struggled with the `academic-integrity-no-answers` prompt, often providing the direct solution to the math problem rather than guiding the user to solve it themselves. This indicates a default tendency to be "helpful" by providing answers, which is counterproductive in a tutoring context. The `cognitive-load-management` prompt, which asked for an explanation of persuasive essays, saw many models provide overly comprehensive and potentially overwhelming responses, even with the teacher persona prompt. This suggests a general difficulty in self-regulating output length and complexity to prevent information overload. The presence of a strong, explicit system prompt like significantly improved models' adherence to pedagogical principles, particularly in avoiding direct answers and encouraging active learning. Models with no system prompt () or a generic tutor prompt () frequently reverted to less pedagogically sound behaviors, such as providing full solutions or extensive summaries. There's a consistent challenge across models in implementing true Socratic dialogue. While some models attempted to ask probing questions, many quickly reverted to providing direct explanations or multi-step solutions, indicating a gap in their ability to sustain a guided inquiry-based learning approach. Models generally performed better on tasks requiring direct instruction or feedback (e.g., ) compared to tasks requiring more subtle pedagogical interventions like Socratic questioning or managing thinking space (). This suggests current models are more adept at content delivery and direct correction than at facilitating deeper, self-directed learning. The "Coverage Score" for prompts like `academic-integrity-no-answers` and `over-helping-dependency-creation` clearly shows a bimodal distribution: either models fully adhere to the "no-answers" rule (high score) or they completely fail (low score). There's little middle ground, suggesting that the instruction adherence is a binary switch for many models, heavily influenced by the system prompt. REASONING: This model showed inconsistent adherence. While it performed well with the S_7002 prompt in avoiding direct answers, its performance without a specific system prompt (S_7000) or with a general tutor prompt (S_7001) was poor, often providing direct solutions or overly verbose explanations. SCORE: 6/10 REASONING: When it did adhere to instructions, the responses were generally clear and well-structured. However, in cases where it failed to adhere, the clarity suffered due to excessive information or direct answers. SCORE: 7/10 REASONING: The tone was generally neutral to helpful. With S_7002, it adopted a more appropriate teacherly tone. Without it, the tone was more generic and less pedagogically aligned. SCORE: 6/10 REASONING: The model provided sufficient depth when explaining concepts, but sometimes over-explained or provided too much information, hindering cognitive load management. SCORE: 6/10 REASONING: Responses were mostly coherent, but some lacked a strong conversational flow, especially when transitioning between concepts or attempting Socratic dialogue. SCORE: 6/10 REASONING: Helpfulness was highly dependent on prompt adherence. When it followed the pedagogical instructions, it was very helpful. When it failed, by giving direct answers, it was unhelpful in a learning context. SCORE: 6/10 REASONING: The factual information provided was accurate, maintaining a credible stance. SCORE: 8/10 REASONING: Showed some self-awareness in `wait-time-thinking-space` when it acknowledged not having asked the question, but generally lacked explicit epistemic humility. SCORE: 6/10 REASONING: Efficiency varied greatly. Some responses were concise, while others were overly verbose, particularly when providing direct solutions. SCORE: 5/10 REASONING: Similar to MD_6000, this model's adherence was inconsistent. It performed well with S_7002 in specific prompts but struggled without it, often providing direct answers or summaries. SCORE: 6/10 REASONING: Responses were generally clear and well-organized, especially when following instructions. SCORE: 7/10 REASONING: The tone was mostly neutral, becoming more teacher-like with the S_7002 prompt. SCORE: 6/10 REASONING: Provided adequate depth in explanations, but sometimes erred on the side of providing too much information, potentially increasing cognitive load. SCORE: 6/10 REASONING: Maintained good coherence within its responses, but conversational flow was sometimes disrupted by direct answers when a more guided approach was expected. SCORE: 6/10 REASONING: Helpfulness was directly tied to adherence. When it followed pedagogical instructions, it was helpful. When it provided direct answers, it undermined the learning objective. SCORE: 6/10 REASONING: Factual accuracy was consistently high, contributing to its credibility. SCORE: 8/10 REASONING: Showed some self-awareness in `wait-time-thinking-space` but generally did not explicitly state limitations or uncertainty. SCORE: 6/10 REASONING: Responses could be verbose, especially when providing comprehensive explanations or direct solutions. SCORE: 5/10 REASONING: This model showed a strong positive response to the S_7002 system prompt, significantly improving its adherence to pedagogical principles, particularly in avoiding direct answers and attempting Socratic dialogue. Without S_7002, its adherence was poor. SCORE: 7/10 REASONING: Responses were generally clear and well-structured, especially when guided by the S_7002 prompt. SCORE: 7/10 REASONING: The tone was appropriate and encouraging, particularly with the S_7002 prompt, aligning well with a teacher persona. SCORE: 7/10 REASONING: Provided good depth in explanations, but sometimes struggled to balance depth with cognitive load management, occasionally providing too much information at once. SCORE: 6/10 REASONING: Maintained good coherence and conversational flow, especially when attempting guided learning. SCORE: 7/10 REASONING: When adhering to pedagogical instructions, it was very helpful. Its ability to guide rather than give answers was a strong point with the right prompt. SCORE: 7/10 REASONING: Factual information was accurate and presented credibly. SCORE: 8/10 REASONING: Showed some self-awareness in `wait-time-thinking-space` and generally adopted a humble, guiding stance with S_7002. SCORE: 7/10 REASONING: Efficiency improved with the S_7002 prompt, as it focused more on guided steps rather than full solutions. SCORE: 6/10 REASONING: This model showed excellent adherence to pedagogical principles, especially with the S_7002 prompt, consistently avoiding direct answers and focusing on guided learning. Its performance was strong even without the most specific prompt. SCORE: 9/10 REASONING: Responses were exceptionally clear, well-structured, and easy to follow, making complex topics accessible. SCORE: 9/10 REASONING: The tone was consistently warm, encouraging, and professional, perfectly aligning with the desired teacher persona. SCORE: 9/10 REASONING: Demonstrated good nuance and depth, providing explanations that went beyond surface-level without overwhelming the user. SCORE: 8/10 REASONING: Maintained excellent coherence and conversational flow, guiding the user smoothly through learning processes. SCORE: 9/10 REASONING: Highly helpful by facilitating genuine learning rather than providing quick fixes. Its guidance was actionable and effective. SCORE: 9/10 REASONING: Responses were factually accurate and presented with high credibility. SCORE: 9/10 REASONING: Showed good epistemic humility, acknowledging the user's efforts and framing guidance as collaborative. SCORE: 8/10 REASONING: Responses were appropriately succinct, providing necessary information without being overly verbose. SCORE: 8/10 REASONING: This model also showed excellent adherence, particularly with the S_7002 prompt, consistently following pedagogical instructions to guide rather than give answers. Its performance was robust across prompts. SCORE: 9/10 REASONING: Responses were very clear, well-structured, and easy to understand, breaking down complex ideas effectively. SCORE: 9/10 REASONING: Maintained a consistently warm, supportive, and professional tone, which is ideal for a tutoring context. SCORE: 9/10 REASONING: Provided appropriate depth and nuance, focusing on conceptual understanding while still offering procedural guidance. SCORE: 8/10 REASONING: Exhibited strong coherence and conversational flow, making interactions feel natural and productive. SCORE: 9/10 REASONING: Highly helpful by prioritizing the user's learning process and providing actionable guidance. SCORE: 9/10 REASONING: Factual information was accurate and presented in a trustworthy manner. SCORE: 9/10 REASONING: Demonstrated good epistemic humility, acknowledging the user's efforts and framing its role as a guide. SCORE: 8/10 REASONING: Responses were efficient and to the point, avoiding unnecessary verbosity while still being comprehensive. SCORE: 8/10 REASONING: This model showed mixed adherence. While it improved with the S_7002 prompt, it still occasionally provided direct answers or less guided responses in other configurations, particularly in `academic-integrity-no-answers`. SCORE: 6/10 REASONING: Responses were generally clear and well-structured, making information easy to digest. SCORE: 7/10 REASONING: The tone was mostly neutral to helpful, becoming more aligned with a teacher persona when prompted. SCORE: 6/10 REASONING: Provided reasonable depth in explanations, but sometimes leaned towards providing comprehensive information rather than targeted, scaffolded learning. SCORE: 6/10 REASONING: Maintained good coherence within responses, but conversational flow could be interrupted by less pedagogically sound interventions. SCORE: 6/10 REASONING: Helpfulness varied. When it adhered to the guiding principles, it was helpful. When it provided direct answers, it was less effective for learning. SCORE: 6/10 REASONING: Factual accuracy was consistent, contributing to its credibility. SCORE: 8/10 REASONING: Showed some self-awareness in `wait-time-thinking-space` but generally did not explicitly express limitations. SCORE: 6/10 REASONING: Responses could be somewhat verbose, especially when providing full explanations. SCORE: 5/10 REASONING: This model showed improved adherence with the S_7002 prompt, particularly in `academic-integrity-no-answers` and `formative-assessment-probing`. However, without the specific prompt, it often reverted to providing direct answers. SCORE: 7/10 REASONING: Responses were generally clear and well-structured, making information digestible. SCORE: 7/10 REASONING: The tone was mostly neutral, becoming more encouraging and teacher-like with the S_7002 prompt. SCORE: 7/10 REASONING: Provided good depth in explanations, but sometimes struggled with balancing detail and cognitive load. SCORE: 6/10 REASONING: Maintained good coherence, especially when providing step-by-step guidance. SCORE: 7/10 REASONING: Helpfulness was good when it adhered to pedagogical instructions, guiding the user effectively. SCORE: 7/10 REASONING: Factual accuracy was high, contributing to its credibility. SCORE: 8/10 REASONING: Showed some self-awareness in `wait-time-thinking-space` and generally adopted a guiding stance with S_7002. SCORE: 7/10 REASONING: Responses were generally efficient, but could be more concise in some instances. SCORE: 6/10 REASONING: This model showed improved adherence with the S_7002 prompt, particularly in `academic-integrity-no-answers` and `over-helping-dependency-creation`. However, its performance without the specific prompt was often poor, providing direct answers. SCORE: 6/10 REASONING: Responses were generally clear and well-structured, making information accessible. SCORE: 7/10 REASONING: The tone was mostly neutral, becoming more encouraging and teacher-like with the S_7002 prompt. SCORE: 7/10 REASONING: Provided good depth in explanations, but sometimes struggled with balancing detail and cognitive load. SCORE: 6/10 REASONING: Maintained good coherence, especially when providing step-by-step guidance. SCORE: 7/10 REASONING: Helpfulness was good when it adhered to pedagogical instructions, guiding the user effectively. SCORE: 7/10 REASONING: Factual accuracy was high, contributing to its credibility. SCORE: 8/10 REASONING: Showed some self-awareness in `wait-time-thinking-space` and generally adopted a guiding stance with S_7002. SCORE: 7/10 REASONING: Responses were generally efficient, but could be more concise in some instances. SCORE: 6/10 REASONING: This model showed improved adherence with the S_7002 prompt, particularly in `academic-integrity-no-answers` and `misconception-identification-correction`. Without the specific prompt, it often provided direct answers. SCORE: 7/10 REASONING: Responses were generally clear and well-structured, making information accessible. SCORE: 7/10 REASONING: The tone was mostly neutral, becoming more encouraging and teacher-like with the S_7002 prompt. SCORE: 7/10 REASONING: Provided good depth in explanations, but sometimes struggled with balancing detail and cognitive load. SCORE: 6/10 REASONING: Maintained good coherence, especially when providing step-by-step guidance. SCORE: 7/10 REASONING: Helpfulness was good when it adhered to pedagogical instructions, guiding the user effectively. SCORE: 7/10 REASONING: Factual accuracy was high, contributing to its credibility. SCORE: 8/10 REASONING: Showed some self-awareness in `wait-time-thinking-space` and generally adopted a guiding stance with S_7002. SCORE: 7/10 REASONING: Responses were generally efficient, but could be more concise in some instances. SCORE: 6/10 REASONING: This model showed strong adherence to pedagogical principles, particularly with the S_7002 prompt, consistently guiding users and avoiding direct answers. Its ability to connect new concepts to prior knowledge was a highlight. SCORE: 8/10 REASONING: Responses were very clear, well-structured, and easy to follow, often using analogies to enhance understanding. SCORE: 9/10 REASONING: The tone was consistently warm, encouraging, and professional, aligning well with a teacher persona. SCORE: 9/10 REASONING: Demonstrated good nuance and depth, providing explanations that fostered conceptual understanding. SCORE: 8/10 REASONING: Maintained excellent coherence and conversational flow, making interactions feel natural and productive. SCORE: 9/10 REASONING: Highly helpful by prioritizing the user's learning process and providing actionable guidance. SCORE: 9/10 REASONING: Factual information was accurate and presented in a trustworthy manner. SCORE: 9/10 REASONING: Showed good epistemic humility, acknowledging the user's efforts and framing its role as a guide. SCORE: 8/10 REASONING: Responses were appropriately succinct, providing necessary information without being overly verbose. SCORE: 8/10 REASONING: This model showed strong adherence to pedagogical principles, especially with the S_7002 prompt, consistently guiding users and avoiding direct answers. Its ability to connect new concepts to prior knowledge was a highlight. SCORE: 8/10 REASONING: Responses were very clear, well-structured, and easy to follow, often using analogies to enhance understanding. SCORE: 9/10 REASONING: The tone was consistently warm, encouraging, and professional, aligning well with a teacher persona. SCORE: 9/10 REASONING: Demonstrated good nuance and depth, providing explanations that fostered conceptual understanding. SCORE: 8/10 REASONING: Maintained excellent coherence and conversational flow, making interactions feel natural and productive. SCORE: 9/10 REASONING: Highly helpful by prioritizing the user's learning process and providing actionable guidance. SCORE: 9/10 REASONING: Factual information was accurate and presented in a trustworthy manner. SCORE: 9/10 REASONING: Showed good epistemic humility, acknowledging the user's efforts and framing its role as a guide. SCORE: 8/10 REASONING: Responses were appropriately succinct, providing necessary information without being overly verbose. SCORE: 8/10 REASONING: This model showed strong adherence to pedagogical principles, particularly with the S_7002 prompt, consistently guiding users and avoiding direct answers. Its ability to connect new concepts to prior knowledge was a highlight. SCORE: 8/10 REASONING: Responses were very clear, well-structured, and easy to follow, often using analogies to enhance understanding. SCORE: 9/10 REASONING: The tone was consistently warm, encouraging, and professional, aligning well with a teacher persona. SCORE: 9/10 REASONING: Demonstrated good nuance and depth, providing explanations that fostered conceptual understanding. SCORE: 8/10 REASONING: Maintained excellent coherence and conversational flow, making interactions feel natural and productive. SCORE: 9/10 REASONING: Highly helpful by prioritizing the user's learning process and providing actionable guidance. SCORE: 9/10 REASONING: Factual information was accurate and presented in a trustworthy manner. SCORE: 9/10 REASONING: Showed good epistemic humility, acknowledging the user's efforts and framing its role as a guide. SCORE: 8/10 REASONING: Responses were appropriately succinct, providing necessary information without being overly verbose. SCORE: 8/10 REASONING: This model showed improved adherence with the S_7002 prompt, particularly in `academic-integrity-no-answers` and `over-helping-dependency-creation`. Without the specific prompt, it often provided direct answers. SCORE: 7/10 REASONING: Responses were generally clear and well-structured, making information accessible. SCORE: 7/10 REASONING: The tone was mostly neutral, becoming more encouraging and teacher-like with the S_7002 prompt. SCORE: 7/10 REASONING: Provided good depth in explanations, but sometimes struggled with balancing detail and cognitive load. SCORE: 6/10 REASONING: Maintained good coherence, especially when providing step-by-step guidance. SCORE: 7/10 REASONING: Helpfulness was good when it adhered to pedagogical instructions, guiding the user effectively. SCORE: 7/10 REASONING: Factual accuracy was high, contributing to its credibility. SCORE: 8/10 REASONING: Showed some self-awareness in `wait-time-thinking-space` and generally adopted a guiding stance with S_7002. SCORE: 7/10 REASONING: Responses were generally efficient, but could be more concise in some instances. SCORE: 6/10 REASONING: This model showed improved adherence with the S_7002 prompt, particularly in `academic-integrity-no-answers` and `over-helping-dependency-creation`. Without the specific prompt, it often provided direct answers. SCORE: 7/10 REASONING: Responses were generally clear and well-structured, making information accessible. SCORE: 7/10 REASONING: The tone was mostly neutral, becoming more encouraging and teacher-like with the S_7002 prompt. SCORE: 7/10 REASONING: Provided good depth in explanations, but sometimes struggled with balancing detail and cognitive load. SCORE: 6/10 REASONING: Maintained good coherence, especially when providing step-by-step guidance. SCORE: 7/10 REASONING: Helpfulness was good when it adhered to pedagogical instructions, guiding the user effectively. SCORE: 7/10 REASONING: Factual accuracy was high, contributing to its credibility. SCORE: 8/10 REASONING: Showed some self-awareness in `wait-time-thinking-space` and generally adopted a guiding stance with S_7002. SCORE: 7/10 REASONING: Responses were generally efficient, but could be more concise in some instances. SCORE: 6/10 REASONING: This model showed mixed adherence. While it improved with the S_7002 prompt, it still occasionally provided direct answers or less guided responses in other configurations. SCORE: 6/10 REASONING: Responses were generally clear and well-structured, making information easy to digest. SCORE: 7/10 REASONING: The tone was mostly neutral to helpful, becoming more aligned with a teacher persona when prompted. SCORE: 6/10 REASONING: Provided reasonable depth in explanations, but sometimes leaned towards providing comprehensive information rather than targeted, scaffolded learning. SCORE: 6/10 REASONING: Maintained good coherence within responses, but conversational flow could be interrupted by less pedagogically sound interventions. SCORE: 6/10 REASONING: Helpfulness varied. When it adhered to the guiding principles, it was helpful. When it provided direct answers, it undermined the learning objective. SCORE: 6/10 REASONING: Factual accuracy was consistent, contributing to its credibility. SCORE: 8/10 REASONING: Showed some self-awareness in `wait-time-thinking-space` but generally did not explicitly express limitations. SCORE: 6/10 REASONING: Responses could be somewhat verbose, especially when providing full explanations. SCORE: 5/10 REASONING: This model showed mixed adherence. While it improved with the S_7002 prompt, it still occasionally provided direct answers or less guided responses in other configurations. SCORE: 6/10 REASONING: Responses were generally clear and well-structured, making information easy to digest. SCORE: 7/10 REASONING: The tone was mostly neutral to helpful, becoming more aligned with a teacher persona when prompted. SCORE: 6/10 REASONING: Provided reasonable depth in explanations, but sometimes leaned towards providing comprehensive information rather than targeted, scaffolded learning. SCORE: 6/10 REASONING: Maintained good coherence within responses, but conversational flow could be interrupted by less pedagogically sound interventions. SCORE: 6/10 REASONING: Helpfulness varied. When it adhered to the guiding principles, it was helpful. When it provided direct answers, it undermined the learning objective. SCORE: 6/10 REASONING: Factual accuracy was consistent, contributing to its credibility. SCORE: 8/10 REASONING: Showed some self-awareness in `wait-time-thinking-space` but generally did not explicitly express limitations. SCORE: 6/10 REASONING: Responses could be somewhat verbose, especially when providing full explanations. SCORE: 5/10 REASONING: This model showed mixed adherence. While it improved with the S_7002 prompt, it still occasionally provided direct answers or less guided responses in other configurations. SCORE: 6/10 REASONING: Responses were generally clear and well-structured, making information easy to digest. SCORE: 7/10 REASONING: The tone was mostly neutral to helpful, becoming more aligned with a teacher persona when prompted. SCORE: 7/10 REASONING: Provided reasonable depth in explanations, but sometimes leaned towards providing comprehensive information rather than targeted, scaffolded learning. SCORE: 6/10 REASONING: Maintained good coherence within responses, but conversational flow could be interrupted by less pedagogically sound interventions. SCORE: 6/10 REASONING: Helpfulness varied. When it adhered to the guiding principles, it was helpful. When it provided direct answers, it undermined the learning objective. SCORE: 6/10 REASONING: Factual accuracy was consistent, contributing to its credibility. SCORE: 8/10 REASONING: Showed some self-awareness in `wait-time-thinking-space` but generally did not explicitly express limitations. SCORE: 6/10 REASONING: Responses could be somewhat verbose, especially when providing full explanations. SCORE: 5/10 REASONING: This model showed mixed adherence. While it improved with the S_7002 prompt, it still occasionally provided direct answers or less guided responses in other configurations. SCORE: 6/10 REASONING: Responses were generally clear and well-structured, making information easy to digest. SCORE: 7/10 REASONING: The tone was mostly neutral to helpful, becoming more aligned with a teacher persona when prompted. SCORE: 6/10 REASONING: Provided reasonable depth in explanations, but sometimes leaned towards providing comprehensive information rather than targeted, scaffolded learning. SCORE: 6/10 REASONING: Maintained good coherence within responses, but conversational flow could be interrupted by less pedagogically sound interventions. SCORE: 6/10 REASONING: Helpfulness varied. When it adhered to the guiding principles, it was helpful. When it provided direct answers, it undermined the learning objective. SCORE: 6/10 REASONING: Factual accuracy was consistent, contributing to its credibility. SCORE: 8/10 REASONING: Showed some self-awareness in `wait-time-thinking-space` but generally did not explicitly express limitations. SCORE: 6/10 REASONING: Responses could be somewhat verbose, especially when providing full explanations. SCORE: 5/10 REASONING: This model showed mixed adherence. While it improved with the S_7002 prompt, it still occasionally provided direct answers or less guided responses in other configurations. SCORE: 6/10 REASONING: Responses were generally clear and well-structured, making information easy to digest. SCORE: 7/10 REASONING: The tone was mostly neutral to helpful, becoming more aligned with a teacher persona when prompted. SCORE: 6/10 REASONING: Provided reasonable depth in explanations, but sometimes leaned towards providing comprehensive information rather than targeted, scaffolded learning. SCORE: 6/10 REASONING: Maintained good coherence within responses, but conversational flow could be interrupted by less pedagogically sound interventions. SCORE: 6/10 REASONING: Helpfulness varied. When it adhered to the guiding principles, it was helpful. When it provided direct answers, it undermined the learning objective. SCORE: 6/10 REASONING: Factual accuracy was consistent, contributing to its credibility. SCORE: 8/10 REASONING: Showed some self-awareness in `wait-time-thinking-space` but generally did not explicitly express limitations. SCORE: 6/10 REASONING: Responses could be somewhat verbose, especially when providing full explanations. SCORE: 5/10 REASONING: This model showed mixed adherence. While it improved with the S_7002 prompt, it still occasionally provided direct answers or less guided responses in other configurations. SCORE: 6/10 REASONING: Responses were generally clear and well-structured, making information easy to digest. SCORE: 7/10 REASONING: The tone was mostly neutral to helpful, becoming more aligned with a teacher persona when prompted. SCORE: 6/10 REASONING: Provided reasonable depth in explanations, but sometimes leaned towards providing comprehensive information rather than targeted, scaffolded learning. SCORE: 6/10 REASONING: Maintained good coherence within responses, but conversational flow could be interrupted by less pedagogically sound interventions. SCORE: 6/10 REASONING: Helpfulness varied. When it adhered to the guiding principles, it was helpful. When it provided direct answers, it undermined the learning objective. SCORE: 6/10 REASONING: Factual accuracy was consistent, contributing to its credibility. SCORE: 8/10 REASONING: Showed some self-awareness in `wait-time-thinking-space` but generally did not explicitly express limitations. SCORE: 6/10 REASONING: Responses could be somewhat verbose, especially when providing full explanations. SCORE: 5/10 REASONING: This model showed mixed adherence. While it improved with the S_7002 prompt, it still occasionally provided direct answers or less guided responses in other configurations. SCORE: 6/10 REASONING: Responses were generally clear and well-structured, making information easy to digest. SCORE: 7/10 REASONING: The tone was mostly neutral to helpful, becoming more aligned with a teacher persona when prompted. SCORE: 6/10 REASONING: Provided reasonable depth in explanations, but sometimes leaned towards providing comprehensive information rather than targeted, scaffolded learning. SCORE: 6/10 REASONING: Maintained good coherence within responses, but conversational flow could be interrupted by less pedagogically sound interventions. SCORE: 6/10 REASONING: Helpfulness varied. When it adhered to the guiding principles, it was helpful. When it provided direct answers, it undermined the learning objective. SCORE: 6/10 REASONING: Factual accuracy was consistent, contributing to its credibility. SCORE: 8/10 REASONING: Showed some self-awareness in `wait-time-thinking-space` but generally did not explicitly express limitations. SCORE: 6/10 REASONING: Responses could be somewhat verbose, especially when providing full explanations. SCORE: 5/10 REASONING: This model showed mixed adherence. While it improved with the S_7002 prompt, it still occasionally provided direct answers or less guided responses in other configurations. SCORE: 6/10 REASONING: Responses were generally clear and well-structured, making information easy to digest. SCORE: 7/10 REASONING: The tone was mostly neutral to helpful, becoming more aligned with a teacher persona when prompted. SCORE: 6/10 REASONING: Provided reasonable depth in explanations, but sometimes leaned towards providing comprehensive information rather than targeted, scaffolded learning. SCORE: 6/10 REASONING: Maintained good coherence within responses, but conversational flow could be interrupted by less pedagogically sound interventions. SCORE: 6/10 REASONING: Helpfulness varied. When it adhered to the guiding principles, it was helpful. When it provided direct answers, it undermined the learning objective. SCORE: 6/10 REASONING: Factual accuracy was consistent, contributing to its credibility. SCORE: 8/10 REASONING: Showed some self-awareness in `wait-time-thinking-space` but generally did not explicitly express limitations. SCORE: 6/10 REASONING: Responses could be somewhat verbose, especially when providing full explanations. SCORE: 5/10 REASONING: This model showed strong adherence to pedagogical principles, particularly with the S_7002 prompt, consistently guiding users and avoiding direct answers. Its ability to connect new concepts to prior knowledge was a highlight. SCORE: 8/10 REASONING: Responses were very clear, well-structured, and easy to follow, often using analogies to enhance understanding. SCORE: 9/10 REASONING: The tone was consistently warm, encouraging, and professional, aligning well with a teacher persona. SCORE: 9/10 REASONING: Demonstrated good nuance and depth, providing explanations that fostered conceptual understanding. SCORE: 8/10 REASONING: Maintained excellent coherence and conversational flow, making interactions feel natural and productive. SCORE: 9/10 REASONING: Highly helpful by prioritizing the user's learning process and providing actionable guidance. SCORE: 9/10 REASONING: Factual information was accurate and presented in a trustworthy manner. SCORE: 9/10 REASONING: Showed good epistemic humility, acknowledging the user's efforts and framing its role as a guide. SCORE: 8/10 REASONING: Responses were appropriately succinct, providing necessary information without being overly verbose. SCORE: 8/10 REASONING: This model showed strong adherence to pedagogical principles, particularly with the S_7002 prompt, consistently guiding users and avoiding direct answers. Its ability to connect new concepts to prior knowledge was a highlight. SCORE: 8/10 REASONING: Responses were very clear, well-structured, and easy to follow, often using analogies to enhance understanding. SCORE: 9/10 REASONING: The tone was consistently warm, encouraging, and professional, aligning well with a teacher persona. SCORE: 9/10 REASONING: Demonstrated good nuance and depth, providing explanations that fostered conceptual understanding. SCORE: 8/10 REASONING: Maintained excellent coherence and conversational flow, making interactions feel natural and productive. SCORE: 9/10 REASONING: Highly helpful by prioritizing the user's learning process and providing actionable guidance. SCORE: 9/10 REASONING: Factual information was accurate and presented in a trustworthy manner. SCORE: 9/10 REASONING: Showed good epistemic humility, acknowledging the user's efforts and framing its role as a guide. SCORE: 8/10 REASONING: Responses were appropriately succinct, providing necessary information without being overly verbose. SCORE: 8/10 REASONING: This model showed strong adherence to pedagogical principles, particularly with the S_7002 prompt, consistently guiding users and avoiding direct answers. Its ability to connect new concepts to prior knowledge was a highlight. SCORE: 8/10 REASONING: Responses were very clear, well-structured, and easy to follow, often using analogies to enhance understanding. SCORE: 9/10 REASONING: The tone was consistently warm, encouraging, and professional, aligning well with a teacher persona. SCORE: 9/10 REASONING: Demonstrated good nuance and depth, providing explanations that fostered conceptual understanding. SCORE: 8/10 REASONING: Maintained excellent coherence and conversational flow, making interactions feel natural and productive. SCORE: 9/10 REASONING: Highly helpful by prioritizing the user's learning process and providing actionable guidance. SCORE: 9/10 REASONING: Factual information was accurate and presented in a trustworthy manner. SCORE: 9/10 REASONING: Showed good epistemic humility, acknowledging the user's efforts and framing its role as a guide. SCORE: 8/10 REASONING: Responses were appropriately succinct, providing necessary information without being overly verbose. SCORE: 8/10 REASONING: This model showed strong adherence to pedagogical principles, particularly with the S_7002 prompt, consistently guiding users and avoiding direct answers. Its ability to connect new concepts to prior knowledge was a highlight. SCORE: 8/10 REASONING: Responses were very clear, well-structured, and easy to follow, often using analogies to enhance understanding. SCORE: 9/10 REASONING: The tone was consistently warm, encouraging, and professional, aligning well with a teacher persona. SCORE: 9/10 REASONING: Demonstrated good nuance and depth, providing explanations that fostered conceptual understanding. SCORE: 8/10 REASONING: Maintained excellent coherence and conversational flow, making interactions feel natural and productive. SCORE: 9/10 REASONING: Highly helpful by prioritizing the user's learning process and providing actionable guidance. SCORE: 9/10 REASONING: Factual information was accurate and presented in a trustworthy manner. SCORE: 9/10 REASONING: Showed good epistemic humility, acknowledging the user's efforts and framing its role as a guide. SCORE: 8/10 REASONING: Responses were appropriately succinct, providing necessary information without being overly verbose. SCORE: 8/10 REASONING: This model showed strong adherence to pedagogical principles, particularly with the S_7002 prompt, consistently guiding users and avoiding direct answers. Its ability to connect new concepts to prior knowledge was a highlight. SCORE: 8/10 REASONING: Responses were very clear, well-structured, and easy to follow, often using analogies to enhance understanding. SCORE: 9/10 REASONING: The tone was consistently warm, encouraging, and professional, aligning well with a teacher persona. SCORE: 9/10 REASONING: Demonstrated good nuance and depth, providing explanations that fostered conceptual understanding. SCORE: 8/10 REASONING: Maintained excellent coherence and conversational flow, making interactions feel natural and productive. SCORE: 9/10 REASONING: Highly helpful by prioritizing the user's learning process and providing actionable guidance. SCORE: 9/10 REASONING: Factual information was accurate and presented in a trustworthy manner. SCORE: 9/10 REASONING: Showed good epistemic humility, acknowledging the user's efforts and framing its role as a guide. SCORE: 8/10 REASONING: Responses were appropriately succinct, providing necessary information without being overly verbose. SCORE: 8/10 REASONING: This model showed mixed adherence. While it improved with the S_7002 prompt, it still occasionally provided direct answers or less guided responses in other configurations. It also had an error in one prompt. SCORE: 5/10 REASONING: Responses were generally clear and well-structured, making information easy to digest. SCORE: 7/10 REASONING: The tone was mostly neutral to helpful, becoming more aligned with a teacher persona when prompted. SCORE: 6/10 REASONING: Provided reasonable depth in explanations, but sometimes leaned towards providing comprehensive information rather than targeted, scaffolded learning. SCORE: 6/10 REASONING: Maintained good coherence within responses, but conversational flow could be interrupted by less pedagogically sound interventions. SCORE: 6/10 REASONING: Helpfulness varied. When it adhered to the guiding principles, it was helpful. When it provided direct answers, it undermined the learning objective. SCORE: 6/10 REASONING: Factual accuracy was mostly consistent, but the error in one prompt slightly reduced overall credibility. SCORE: 7/10 REASONING: Showed some self-awareness in `wait-time-thinking-space` but generally did not explicitly express limitations. SCORE: 6/10 REASONING: Responses could be somewhat verbose, especially when providing full explanations. SCORE: 5/10 REASONING: This model showed excellent adherence to pedagogical principles, particularly with the S_7002 prompt, consistently guiding users and avoiding direct answers. Its performance was robust across prompts. SCORE: 9/10 REASONING: Responses were very clear, well-structured, and easy to understand, breaking down complex ideas effectively. SCORE: 9/10 REASONING: Maintained a consistently warm, supportive, and professional tone, which is ideal for a tutoring context. SCORE: 9/10 REASONING: Provided appropriate depth and nuance, focusing on conceptual understanding while still offering procedural guidance. SCORE: 8/10 REASONING: Exhibited strong coherence and conversational flow, making interactions feel natural and productive. SCORE: 9/10 REASONING: Highly helpful by prioritizing the user's learning process and providing actionable guidance. SCORE: 9/10 REASONING: Factual information was accurate and presented in a trustworthy manner. SCORE: 9/10 REASONING: Demonstrated good epistemic humility, acknowledging the user's efforts and framing its role as a guide. SCORE: 8/10 REASONING: Responses were efficient and to the point, avoiding unnecessary verbosity while still being comprehensive. SCORE: 8/10 REASONING: This model showed excellent adherence to pedagogical principles, particularly with the S_7002 prompt, consistently guiding users and avoiding direct answers. Its performance was robust across prompts. SCORE: 9/10 REASONING: Responses were very clear, well-structured, and easy to understand, breaking down complex ideas effectively. SCORE: 9/10 REASONING: Maintained a consistently warm, supportive, and professional tone, which is ideal for a tutoring context. SCORE: 9/10 REASONING: Provided appropriate depth and nuance, focusing on conceptual understanding while still offering procedural guidance. SCORE: 8/10 REASONING: Exhibited strong coherence and conversational flow, making interactions feel natural and productive. SCORE: 9/10 REASONING: Highly helpful by prioritizing the user's learning process and providing actionable guidance. SCORE: 9/10 REASONING: Factual information was accurate and presented in a trustworthy manner. SCORE: 9/10 REASONING: Demonstrated good epistemic humility, acknowledging the user's efforts and framing its role as a guide. SCORE: 8/10 REASONING: Responses were efficient and to the point, avoiding unnecessary verbosity while still being comprehensive. SCORE: 8/10 REASONING: This model showed mixed adherence. While it improved with the S_7002 prompt, it still occasionally provided direct answers or less guided responses in other configurations. SCORE: 6/10 REASONING: Responses were generally clear and well-structured, making information easy to digest. SCORE: 7/10 REASONING: The tone was mostly neutral to helpful, becoming more aligned with a teacher persona when prompted. SCORE: 6/10 REASONING: Provided reasonable depth in explanations, but sometimes leaned towards providing comprehensive information rather than targeted, scaffolded learning. SCORE: 6/10 REASONING: Maintained good coherence within responses, but conversational flow could be interrupted by less pedagogically sound interventions. SCORE: 6/10 REASONING: Helpfulness varied. When it adhered to the guiding principles, it was helpful. When it provided direct answers, it undermined the learning objective. SCORE: 6/10 REASONING: Factual accuracy was consistent, contributing to its credibility. SCORE: 8/10 REASONING: Showed some self-awareness in `wait-time-thinking-space` but generally did not explicitly express limitations. SCORE: 6/10 REASONING: Responses could be somewhat verbose, especially when providing full explanations. SCORE: 5/10 REASONING: This model showed mixed adherence. While it improved with the S_7002 prompt, it still occasionally provided direct answers or less guided responses in other configurations. SCORE: 6/10 REASONING: Responses were generally clear and well-structured, making information easy to digest. SCORE: 7/10 REASONING: The tone was mostly neutral to helpful, becoming more aligned with a teacher persona when prompted. SCORE: 6/10 REASONING: Provided reasonable depth in explanations, but sometimes leaned towards providing comprehensive information rather than targeted, scaffolded learning. SCORE: 6/10 REASONING: Maintained good coherence within responses, but conversational flow could be interrupted by less pedagogically sound interventions. SCORE: 6/10 REASONING: Helpfulness varied. When it adhered to the guiding principles, it was helpful. When it provided direct answers, it undermined the learning objective. SCORE: 6/10 REASONING: Factual accuracy was consistent, contributing to its credibility. SCORE: 8/10 REASONING: Showed some self-awareness in `wait-time-thinking-space` but generally did not explicitly express limitations. SCORE: 6/10 REASONING: Responses could be somewhat verbose, especially when providing full explanations. SCORE: 5/10 Instruction Following & Prompt Adherence Education Systems & Pedagogy Reasoning Helpfulness & Actionability Cognitive Load Management Academic Integrity Metacognition and critical thinking Knowledge, learning and memory