| {"id": "topic-00-style-0", "topic_id": 0, "split": "train", "prompt": "Explain root mean square of an acceleration waveform in three concise sentences for a junior engineer.", "source": "original Sofia prompt bank v1", "response": "RMS is sqrt(mean(x**2)) for sampled acceleration x. It measures signal magnitude, not the signed mean; a sine wave of amplitude A has RMS A/sqrt(2). State the acceleration unit and the time interval used.", "teacher": "Qwen/Qwen2.5-0.5B-Instruct", "teacher_revision": "7ae557604adf67be50417f59c2c2f167def9a775"} | |
| {"id": "topic-00-style-1", "topic_id": 0, "split": "train", "prompt": "What is root mean square of an acceleration waveform? Give a short practical example.", "source": "original Sofia prompt bank v1", "response": "RMS is sqrt(mean(x**2)) for sampled acceleration x. It measures signal magnitude, not the signed mean; a sine wave of amplitude A has RMS A/sqrt(2). State the acceleration unit and the time interval used.", "teacher": "Qwen/Qwen2.5-0.5B-Instruct", "teacher_revision": "7ae557604adf67be50417f59c2c2f167def9a775"} | |
| {"id": "topic-00-style-2", "topic_id": 0, "split": "train", "prompt": "Describe one common mistake involving root mean square of an acceleration waveform and how to avoid it.", "source": "original Sofia prompt bank v1", "response": "RMS is sqrt(mean(x**2)) for sampled acceleration x. It measures signal magnitude, not the signed mean; a sine wave of amplitude A has RMS A/sqrt(2). State the acceleration unit and the time interval used.", "teacher": "Qwen/Qwen2.5-0.5B-Instruct", "teacher_revision": "7ae557604adf67be50417f59c2c2f167def9a775"} | |
| {"id": "topic-00-style-3", "topic_id": 0, "split": "train", "prompt": "Give a brief technical explanation of root mean square of an acceleration waveform. State an important limitation.", "source": "original Sofia prompt bank v1", "response": "RMS is sqrt(mean(x**2)) for sampled acceleration x. It measures signal magnitude, not the signed mean; a sine wave of amplitude A has RMS A/sqrt(2). State the acceleration unit and the time interval used.", "teacher": "Qwen/Qwen2.5-0.5B-Instruct", "teacher_revision": "7ae557604adf67be50417f59c2c2f167def9a775"} | |
| {"id": "topic-01-style-0", "topic_id": 1, "split": "train", "prompt": "Explain crest factor and impulsive vibration in three concise sentences for a junior engineer.", "source": "original Sofia prompt bank v1", "response": "Crest factor is max(abs(x))/RMS(x). A large value can indicate sparse impulses, but noise, saturation and window duration also affect it. It does not identify a bearing fault by itself.", "teacher": "Qwen/Qwen2.5-0.5B-Instruct", "teacher_revision": "7ae557604adf67be50417f59c2c2f167def9a775"} | |
| {"id": "topic-01-style-1", "topic_id": 1, "split": "train", "prompt": "What is crest factor and impulsive vibration? Give a short practical example.", "source": "original Sofia prompt bank v1", "response": "Crest factor is max(abs(x))/RMS(x). A large value can indicate sparse impulses, but noise, saturation and window duration also affect it. It does not identify a bearing fault by itself.", "teacher": "Qwen/Qwen2.5-0.5B-Instruct", "teacher_revision": "7ae557604adf67be50417f59c2c2f167def9a775"} | |
| {"id": "topic-01-style-2", "topic_id": 1, "split": "train", "prompt": "Describe one common mistake involving crest factor and impulsive vibration and how to avoid it.", "source": "original Sofia prompt bank v1", "response": "Crest factor is max(abs(x))/RMS(x). A large value can indicate sparse impulses, but noise, saturation and window duration also affect it. It does not identify a bearing fault by itself.", "teacher": "Qwen/Qwen2.5-0.5B-Instruct", "teacher_revision": "7ae557604adf67be50417f59c2c2f167def9a775"} | |
| {"id": "topic-01-style-3", "topic_id": 1, "split": "train", "prompt": "Give a brief technical explanation of crest factor and impulsive vibration. State an important limitation.", "source": "original Sofia prompt bank v1", "response": "Crest factor is max(abs(x))/RMS(x). A large value can indicate sparse impulses, but noise, saturation and window duration also affect it. It does not identify a bearing fault by itself.", "teacher": "Qwen/Qwen2.5-0.5B-Instruct", "teacher_revision": "7ae557604adf67be50417f59c2c2f167def9a775"} | |
| {"id": "topic-02-style-0", "topic_id": 2, "split": "train", "prompt": "Explain kurtosis and its limitations for fault detection in three concise sentences for a junior engineer.", "source": "original Sofia prompt bank v1", "response": "Pearson kurtosis is mean((x-mean(x))**4)/variance(x)**2. A Gaussian distribution has kurtosis 3; excess kurtosis subtracts 3. Large values suggest heavy tails or impulses, but do not establish the physical cause.", "teacher": "Qwen/Qwen2.5-0.5B-Instruct", "teacher_revision": "7ae557604adf67be50417f59c2c2f167def9a775"} | |
| {"id": "topic-02-style-1", "topic_id": 2, "split": "train", "prompt": "What is kurtosis and its limitations for fault detection? Give a short practical example.", "source": "original Sofia prompt bank v1", "response": "Pearson kurtosis is mean((x-mean(x))**4)/variance(x)**2. A Gaussian distribution has kurtosis 3; excess kurtosis subtracts 3. Large values suggest heavy tails or impulses, but do not establish the physical cause.", "teacher": "Qwen/Qwen2.5-0.5B-Instruct", "teacher_revision": "7ae557604adf67be50417f59c2c2f167def9a775"} | |
| {"id": "topic-02-style-2", "topic_id": 2, "split": "train", "prompt": "Describe one common mistake involving kurtosis and its limitations for fault detection and how to avoid it.", "source": "original Sofia prompt bank v1", "response": "Pearson kurtosis is mean((x-mean(x))**4)/variance(x)**2. A Gaussian distribution has kurtosis 3; excess kurtosis subtracts 3. Large values suggest heavy tails or impulses, but do not establish the physical cause.", "teacher": "Qwen/Qwen2.5-0.5B-Instruct", "teacher_revision": "7ae557604adf67be50417f59c2c2f167def9a775"} | |
| {"id": "topic-02-style-3", "topic_id": 2, "split": "train", "prompt": "Give a brief technical explanation of kurtosis and its limitations for fault detection. State an important limitation.", "source": "original Sofia prompt bank v1", "response": "Pearson kurtosis is mean((x-mean(x))**4)/variance(x)**2. A Gaussian distribution has kurtosis 3; excess kurtosis subtracts 3. Large values suggest heavy tails or impulses, but do not establish the physical cause.", "teacher": "Qwen/Qwen2.5-0.5B-Instruct", "teacher_revision": "7ae557604adf67be50417f59c2c2f167def9a775"} | |
| {"id": "topic-03-style-0", "topic_id": 3, "split": "train", "prompt": "Explain the Nyquist sampling criterion and aliasing in three concise sentences for a junior engineer.", "source": "original Sofia prompt bank v1", "response": "A band-limited signal requires a sample rate greater than twice its highest frequency for ideal reconstruction. An analog anti-alias filter must attenuate components above Nyquist before sampling. Aliased energy cannot generally be removed afterward using a digital filter.", "teacher": "Qwen/Qwen2.5-0.5B-Instruct", "teacher_revision": "7ae557604adf67be50417f59c2c2f167def9a775"} | |
| {"id": "topic-03-style-1", "topic_id": 3, "split": "train", "prompt": "What is the Nyquist sampling criterion and aliasing? Give a short practical example.", "source": "original Sofia prompt bank v1", "response": "A band-limited signal requires a sample rate greater than twice its highest frequency for ideal reconstruction. An analog anti-alias filter must attenuate components above Nyquist before sampling. Aliased energy cannot generally be removed afterward using a digital filter.", "teacher": "Qwen/Qwen2.5-0.5B-Instruct", "teacher_revision": "7ae557604adf67be50417f59c2c2f167def9a775"} | |
| {"id": "topic-03-style-2", "topic_id": 3, "split": "train", "prompt": "Describe one common mistake involving the Nyquist sampling criterion and aliasing and how to avoid it.", "source": "original Sofia prompt bank v1", "response": "A band-limited signal requires a sample rate greater than twice its highest frequency for ideal reconstruction. An analog anti-alias filter must attenuate components above Nyquist before sampling. Aliased energy cannot generally be removed afterward using a digital filter.", "teacher": "Qwen/Qwen2.5-0.5B-Instruct", "teacher_revision": "7ae557604adf67be50417f59c2c2f167def9a775"} | |
| {"id": "topic-03-style-3", "topic_id": 3, "split": "train", "prompt": "Give a brief technical explanation of the Nyquist sampling criterion and aliasing. State an important limitation.", "source": "original Sofia prompt bank v1", "response": "A band-limited signal requires a sample rate greater than twice its highest frequency for ideal reconstruction. An analog anti-alias filter must attenuate components above Nyquist before sampling. Aliased energy cannot generally be removed afterward using a digital filter.", "teacher": "Qwen/Qwen2.5-0.5B-Instruct", "teacher_revision": "7ae557604adf67be50417f59c2c2f167def9a775"} | |
| {"id": "topic-04-style-0", "topic_id": 4, "split": "validation", "prompt": "Explain FFT resolution and window duration in three concise sentences for a junior engineer.", "source": "original Sofia prompt bank v1", "response": "FFT bin spacing is sample_rate/N, or 1/window_duration. A longer measured window gives finer frequency spacing. Zero padding interpolates the spectrum without creating new physical frequency resolution.", "teacher": "Qwen/Qwen2.5-0.5B-Instruct", "teacher_revision": "7ae557604adf67be50417f59c2c2f167def9a775"} | |
| {"id": "topic-04-style-1", "topic_id": 4, "split": "validation", "prompt": "What is FFT resolution and window duration? Give a short practical example.", "source": "original Sofia prompt bank v1", "response": "FFT bin spacing is sample_rate/N, or 1/window_duration. A longer measured window gives finer frequency spacing. Zero padding interpolates the spectrum without creating new physical frequency resolution.", "teacher": "Qwen/Qwen2.5-0.5B-Instruct", "teacher_revision": "7ae557604adf67be50417f59c2c2f167def9a775"} | |
| {"id": "topic-04-style-2", "topic_id": 4, "split": "validation", "prompt": "Describe one common mistake involving FFT resolution and window duration and how to avoid it.", "source": "original Sofia prompt bank v1", "response": "FFT bin spacing is sample_rate/N, or 1/window_duration. A longer measured window gives finer frequency spacing. Zero padding interpolates the spectrum without creating new physical frequency resolution.", "teacher": "Qwen/Qwen2.5-0.5B-Instruct", "teacher_revision": "7ae557604adf67be50417f59c2c2f167def9a775"} | |
| {"id": "topic-04-style-3", "topic_id": 4, "split": "validation", "prompt": "Give a brief technical explanation of FFT resolution and window duration. State an important limitation.", "source": "original Sofia prompt bank v1", "response": "FFT bin spacing is sample_rate/N, or 1/window_duration. A longer measured window gives finer frequency spacing. Zero padding interpolates the spectrum without creating new physical frequency resolution.", "teacher": "Qwen/Qwen2.5-0.5B-Instruct", "teacher_revision": "7ae557604adf67be50417f59c2c2f167def9a775"} | |
| {"id": "topic-05-style-0", "topic_id": 5, "split": "train", "prompt": "Explain spectral leakage and Hann windows in three concise sentences for a junior engineer.", "source": "original Sofia prompt bank v1", "response": "Spectral leakage occurs when a finite observation spreads a tone across frequency bins. A Hann window reduces sidelobes but broadens the main lobe and changes amplitude gain. Apply the appropriate window normalization before comparing amplitudes.", "teacher": "Qwen/Qwen2.5-0.5B-Instruct", "teacher_revision": "7ae557604adf67be50417f59c2c2f167def9a775"} | |
| {"id": "topic-05-style-1", "topic_id": 5, "split": "train", "prompt": "What is spectral leakage and Hann windows? Give a short practical example.", "source": "original Sofia prompt bank v1", "response": "Spectral leakage occurs when a finite observation spreads a tone across frequency bins. A Hann window reduces sidelobes but broadens the main lobe and changes amplitude gain. Apply the appropriate window normalization before comparing amplitudes.", "teacher": "Qwen/Qwen2.5-0.5B-Instruct", "teacher_revision": "7ae557604adf67be50417f59c2c2f167def9a775"} | |
| {"id": "topic-05-style-2", "topic_id": 5, "split": "train", "prompt": "Describe one common mistake involving spectral leakage and Hann windows and how to avoid it.", "source": "original Sofia prompt bank v1", "response": "Spectral leakage occurs when a finite observation spreads a tone across frequency bins. A Hann window reduces sidelobes but broadens the main lobe and changes amplitude gain. Apply the appropriate window normalization before comparing amplitudes.", "teacher": "Qwen/Qwen2.5-0.5B-Instruct", "teacher_revision": "7ae557604adf67be50417f59c2c2f167def9a775"} | |
| {"id": "topic-05-style-3", "topic_id": 5, "split": "train", "prompt": "Give a brief technical explanation of spectral leakage and Hann windows. State an important limitation.", "source": "original Sofia prompt bank v1", "response": "Spectral leakage occurs when a finite observation spreads a tone across frequency bins. A Hann window reduces sidelobes but broadens the main lobe and changes amplitude gain. Apply the appropriate window normalization before comparing amplitudes.", "teacher": "Qwen/Qwen2.5-0.5B-Instruct", "teacher_revision": "7ae557604adf67be50417f59c2c2f167def9a775"} | |
| {"id": "topic-06-style-0", "topic_id": 6, "split": "test", "prompt": "Explain Welch power spectral density estimation in three concise sentences for a junior engineer.", "source": "original Sofia prompt bank v1", "response": "Welch PSD averages periodograms of overlapping windowed segments. Averaging reduces estimator variance at the cost of segment-limited frequency resolution. Use power-per-hertz units and correct sampling-rate and window-energy normalization.", "teacher": "Qwen/Qwen2.5-0.5B-Instruct", "teacher_revision": "7ae557604adf67be50417f59c2c2f167def9a775"} | |
| {"id": "topic-06-style-1", "topic_id": 6, "split": "test", "prompt": "What is Welch power spectral density estimation? Give a short practical example.", "source": "original Sofia prompt bank v1", "response": "Welch PSD averages periodograms of overlapping windowed segments. Averaging reduces estimator variance at the cost of segment-limited frequency resolution. Use power-per-hertz units and correct sampling-rate and window-energy normalization.", "teacher": "Qwen/Qwen2.5-0.5B-Instruct", "teacher_revision": "7ae557604adf67be50417f59c2c2f167def9a775"} | |
| {"id": "topic-06-style-2", "topic_id": 6, "split": "test", "prompt": "Describe one common mistake involving Welch power spectral density estimation and how to avoid it.", "source": "original Sofia prompt bank v1", "response": "Welch PSD averages periodograms of overlapping windowed segments. Averaging reduces estimator variance at the cost of segment-limited frequency resolution. Use power-per-hertz units and correct sampling-rate and window-energy normalization.", "teacher": "Qwen/Qwen2.5-0.5B-Instruct", "teacher_revision": "7ae557604adf67be50417f59c2c2f167def9a775"} | |
| {"id": "topic-06-style-3", "topic_id": 6, "split": "test", "prompt": "Give a brief technical explanation of Welch power spectral density estimation. State an important limitation.", "source": "original Sofia prompt bank v1", "response": "Welch PSD averages periodograms of overlapping windowed segments. Averaging reduces estimator variance at the cost of segment-limited frequency resolution. Use power-per-hertz units and correct sampling-rate and window-energy normalization.", "teacher": "Qwen/Qwen2.5-0.5B-Instruct", "teacher_revision": "7ae557604adf67be50417f59c2c2f167def9a775"} | |
| {"id": "topic-07-style-0", "topic_id": 7, "split": "train", "prompt": "Explain Hilbert envelope analysis for bearing impulses in three concise sentences for a junior engineer.", "source": "original Sofia prompt bank v1", "response": "Envelope analysis forms an analytic signal and examines its amplitude envelope. A selected resonance band can expose repeated modulation from impulses. Filter choice and speed variation matter; an envelope peak alone is not a confirmed bearing defect.", "teacher": "Qwen/Qwen2.5-0.5B-Instruct", "teacher_revision": "7ae557604adf67be50417f59c2c2f167def9a775"} | |
| {"id": "topic-07-style-1", "topic_id": 7, "split": "train", "prompt": "What is Hilbert envelope analysis for bearing impulses? Give a short practical example.", "source": "original Sofia prompt bank v1", "response": "Envelope analysis forms an analytic signal and examines its amplitude envelope. A selected resonance band can expose repeated modulation from impulses. Filter choice and speed variation matter; an envelope peak alone is not a confirmed bearing defect.", "teacher": "Qwen/Qwen2.5-0.5B-Instruct", "teacher_revision": "7ae557604adf67be50417f59c2c2f167def9a775"} | |
| {"id": "topic-07-style-2", "topic_id": 7, "split": "train", "prompt": "Describe one common mistake involving Hilbert envelope analysis for bearing impulses and how to avoid it.", "source": "original Sofia prompt bank v1", "response": "Envelope analysis forms an analytic signal and examines its amplitude envelope. A selected resonance band can expose repeated modulation from impulses. Filter choice and speed variation matter; an envelope peak alone is not a confirmed bearing defect.", "teacher": "Qwen/Qwen2.5-0.5B-Instruct", "teacher_revision": "7ae557604adf67be50417f59c2c2f167def9a775"} | |
| {"id": "topic-07-style-3", "topic_id": 7, "split": "train", "prompt": "Give a brief technical explanation of Hilbert envelope analysis for bearing impulses. State an important limitation.", "source": "original Sofia prompt bank v1", "response": "Envelope analysis forms an analytic signal and examines its amplitude envelope. A selected resonance band can expose repeated modulation from impulses. Filter choice and speed variation matter; an envelope peak alone is not a confirmed bearing defect.", "teacher": "Qwen/Qwen2.5-0.5B-Instruct", "teacher_revision": "7ae557604adf67be50417f59c2c2f167def9a775"} | |
| {"id": "topic-08-style-0", "topic_id": 8, "split": "train", "prompt": "Explain shaft-speed normalization in vibration analysis in three concise sentences for a junior engineer.", "source": "original Sofia prompt bank v1", "response": "Order tracking expresses frequency relative to shaft rotation frequency. A tone at twice shaft speed is order 2. Variable-speed data need suitable resampling or tracking; simply dividing one FFT axis by an average speed may smear features.", "teacher": "Qwen/Qwen2.5-0.5B-Instruct", "teacher_revision": "7ae557604adf67be50417f59c2c2f167def9a775"} | |
| {"id": "topic-08-style-1", "topic_id": 8, "split": "train", "prompt": "What is shaft-speed normalization in vibration analysis? Give a short practical example.", "source": "original Sofia prompt bank v1", "response": "Order tracking expresses frequency relative to shaft rotation frequency. A tone at twice shaft speed is order 2. Variable-speed data need suitable resampling or tracking; simply dividing one FFT axis by an average speed may smear features.", "teacher": "Qwen/Qwen2.5-0.5B-Instruct", "teacher_revision": "7ae557604adf67be50417f59c2c2f167def9a775"} | |
| {"id": "topic-08-style-2", "topic_id": 8, "split": "train", "prompt": "Describe one common mistake involving shaft-speed normalization in vibration analysis and how to avoid it.", "source": "original Sofia prompt bank v1", "response": "Order tracking expresses frequency relative to shaft rotation frequency. A tone at twice shaft speed is order 2. Variable-speed data need suitable resampling or tracking; simply dividing one FFT axis by an average speed may smear features.", "teacher": "Qwen/Qwen2.5-0.5B-Instruct", "teacher_revision": "7ae557604adf67be50417f59c2c2f167def9a775"} | |
| {"id": "topic-08-style-3", "topic_id": 8, "split": "train", "prompt": "Give a brief technical explanation of shaft-speed normalization in vibration analysis. State an important limitation.", "source": "original Sofia prompt bank v1", "response": "Order tracking expresses frequency relative to shaft rotation frequency. A tone at twice shaft speed is order 2. Variable-speed data need suitable resampling or tracking; simply dividing one FFT axis by an average speed may smear features.", "teacher": "Qwen/Qwen2.5-0.5B-Instruct", "teacher_revision": "7ae557604adf67be50417f59c2c2f167def9a775"} | |
| {"id": "topic-09-style-0", "topic_id": 9, "split": "train", "prompt": "Explain the distinction between imbalance and misalignment signatures in three concise sentences for a junior engineer.", "source": "original Sofia prompt bank v1", "response": "Imbalance often emphasizes the first shaft harmonic, while misalignment can produce higher harmonics and axial vibration. These are tendencies, not unique signatures. Mounting, load, machine geometry and physical inspection are needed for diagnosis.", "teacher": "Qwen/Qwen2.5-0.5B-Instruct", "teacher_revision": "7ae557604adf67be50417f59c2c2f167def9a775"} | |
| {"id": "topic-09-style-1", "topic_id": 9, "split": "train", "prompt": "What is the distinction between imbalance and misalignment signatures? Give a short practical example.", "source": "original Sofia prompt bank v1", "response": "Imbalance often emphasizes the first shaft harmonic, while misalignment can produce higher harmonics and axial vibration. These are tendencies, not unique signatures. Mounting, load, machine geometry and physical inspection are needed for diagnosis.", "teacher": "Qwen/Qwen2.5-0.5B-Instruct", "teacher_revision": "7ae557604adf67be50417f59c2c2f167def9a775"} | |
| {"id": "topic-09-style-2", "topic_id": 9, "split": "train", "prompt": "Describe one common mistake involving the distinction between imbalance and misalignment signatures and how to avoid it.", "source": "original Sofia prompt bank v1", "response": "Imbalance often emphasizes the first shaft harmonic, while misalignment can produce higher harmonics and axial vibration. These are tendencies, not unique signatures. Mounting, load, machine geometry and physical inspection are needed for diagnosis.", "teacher": "Qwen/Qwen2.5-0.5B-Instruct", "teacher_revision": "7ae557604adf67be50417f59c2c2f167def9a775"} | |
| {"id": "topic-09-style-3", "topic_id": 9, "split": "train", "prompt": "Give a brief technical explanation of the distinction between imbalance and misalignment signatures. State an important limitation.", "source": "original Sofia prompt bank v1", "response": "Imbalance often emphasizes the first shaft harmonic, while misalignment can produce higher harmonics and axial vibration. These are tendencies, not unique signatures. Mounting, load, machine geometry and physical inspection are needed for diagnosis.", "teacher": "Qwen/Qwen2.5-0.5B-Instruct", "teacher_revision": "7ae557604adf67be50417f59c2c2f167def9a775"} | |
| {"id": "topic-10-style-0", "topic_id": 10, "split": "train", "prompt": "Explain why a synthetic fault classifier needs real-world validation in three concise sentences for a junior engineer.", "source": "original Sofia prompt bank v1", "response": "Synthetic data test a specified generator rather than the full distribution of real machines. High simulated accuracy can reflect easy class formulas or missing confounders. Evaluate independent recorded signals, machine-level splits and acquisition conditions before deployment.", "teacher": "Qwen/Qwen2.5-0.5B-Instruct", "teacher_revision": "7ae557604adf67be50417f59c2c2f167def9a775"} | |
| {"id": "topic-10-style-1", "topic_id": 10, "split": "train", "prompt": "What is why a synthetic fault classifier needs real-world validation? Give a short practical example.", "source": "original Sofia prompt bank v1", "response": "Synthetic data test a specified generator rather than the full distribution of real machines. High simulated accuracy can reflect easy class formulas or missing confounders. Evaluate independent recorded signals, machine-level splits and acquisition conditions before deployment.", "teacher": "Qwen/Qwen2.5-0.5B-Instruct", "teacher_revision": "7ae557604adf67be50417f59c2c2f167def9a775"} | |
| {"id": "topic-10-style-2", "topic_id": 10, "split": "train", "prompt": "Describe one common mistake involving why a synthetic fault classifier needs real-world validation and how to avoid it.", "source": "original Sofia prompt bank v1", "response": "Synthetic data test a specified generator rather than the full distribution of real machines. High simulated accuracy can reflect easy class formulas or missing confounders. Evaluate independent recorded signals, machine-level splits and acquisition conditions before deployment.", "teacher": "Qwen/Qwen2.5-0.5B-Instruct", "teacher_revision": "7ae557604adf67be50417f59c2c2f167def9a775"} | |
| {"id": "topic-10-style-3", "topic_id": 10, "split": "train", "prompt": "Give a brief technical explanation of why a synthetic fault classifier needs real-world validation. State an important limitation.", "source": "original Sofia prompt bank v1", "response": "Synthetic data test a specified generator rather than the full distribution of real machines. High simulated accuracy can reflect easy class formulas or missing confounders. Evaluate independent recorded signals, machine-level splits and acquisition conditions before deployment.", "teacher": "Qwen/Qwen2.5-0.5B-Instruct", "teacher_revision": "7ae557604adf67be50417f59c2c2f167def9a775"} | |
| {"id": "topic-11-style-0", "topic_id": 11, "split": "train", "prompt": "Explain sensor mounting and accelerometer calibration in three concise sentences for a junior engineer.", "source": "original Sofia prompt bank v1", "response": "Accelerometer mounting changes the measured transfer function and usable bandwidth. Calibration relates sensor output to a known physical acceleration. Record orientation, mounting, units, sample rate and clipping status before comparing measurements.", "teacher": "Qwen/Qwen2.5-0.5B-Instruct", "teacher_revision": "7ae557604adf67be50417f59c2c2f167def9a775"} | |
| {"id": "topic-11-style-1", "topic_id": 11, "split": "train", "prompt": "What is sensor mounting and accelerometer calibration? Give a short practical example.", "source": "original Sofia prompt bank v1", "response": "Accelerometer mounting changes the measured transfer function and usable bandwidth. Calibration relates sensor output to a known physical acceleration. Record orientation, mounting, units, sample rate and clipping status before comparing measurements.", "teacher": "Qwen/Qwen2.5-0.5B-Instruct", "teacher_revision": "7ae557604adf67be50417f59c2c2f167def9a775"} | |
| {"id": "topic-11-style-2", "topic_id": 11, "split": "train", "prompt": "Describe one common mistake involving sensor mounting and accelerometer calibration and how to avoid it.", "source": "original Sofia prompt bank v1", "response": "Accelerometer mounting changes the measured transfer function and usable bandwidth. Calibration relates sensor output to a known physical acceleration. Record orientation, mounting, units, sample rate and clipping status before comparing measurements.", "teacher": "Qwen/Qwen2.5-0.5B-Instruct", "teacher_revision": "7ae557604adf67be50417f59c2c2f167def9a775"} | |
| {"id": "topic-11-style-3", "topic_id": 11, "split": "train", "prompt": "Give a brief technical explanation of sensor mounting and accelerometer calibration. State an important limitation.", "source": "original Sofia prompt bank v1", "response": "Accelerometer mounting changes the measured transfer function and usable bandwidth. Calibration relates sensor output to a known physical acceleration. Record orientation, mounting, units, sample rate and clipping status before comparing measurements.", "teacher": "Qwen/Qwen2.5-0.5B-Instruct", "teacher_revision": "7ae557604adf67be50417f59c2c2f167def9a775"} | |
| {"id": "topic-12-style-0", "topic_id": 12, "split": "validation", "prompt": "Explain the difference between acceleration, velocity, and displacement in three concise sentences for a junior engineer.", "source": "original Sofia prompt bank v1", "response": "Acceleration is the time derivative of velocity, and velocity is the derivative of displacement. For a sinusoid, their amplitude factors differ by angular frequency. Integration needs careful treatment of offsets and low-frequency drift.", "teacher": "Qwen/Qwen2.5-0.5B-Instruct", "teacher_revision": "7ae557604adf67be50417f59c2c2f167def9a775"} | |
| {"id": "topic-12-style-1", "topic_id": 12, "split": "validation", "prompt": "What is the difference between acceleration, velocity, and displacement? Give a short practical example.", "source": "original Sofia prompt bank v1", "response": "Acceleration is the time derivative of velocity, and velocity is the derivative of displacement. For a sinusoid, their amplitude factors differ by angular frequency. Integration needs careful treatment of offsets and low-frequency drift.", "teacher": "Qwen/Qwen2.5-0.5B-Instruct", "teacher_revision": "7ae557604adf67be50417f59c2c2f167def9a775"} | |
| {"id": "topic-12-style-2", "topic_id": 12, "split": "validation", "prompt": "Describe one common mistake involving the difference between acceleration, velocity, and displacement and how to avoid it.", "source": "original Sofia prompt bank v1", "response": "Acceleration is the time derivative of velocity, and velocity is the derivative of displacement. For a sinusoid, their amplitude factors differ by angular frequency. Integration needs careful treatment of offsets and low-frequency drift.", "teacher": "Qwen/Qwen2.5-0.5B-Instruct", "teacher_revision": "7ae557604adf67be50417f59c2c2f167def9a775"} | |
| {"id": "topic-12-style-3", "topic_id": 12, "split": "validation", "prompt": "Give a brief technical explanation of the difference between acceleration, velocity, and displacement. State an important limitation.", "source": "original Sofia prompt bank v1", "response": "Acceleration is the time derivative of velocity, and velocity is the derivative of displacement. For a sinusoid, their amplitude factors differ by angular frequency. Integration needs careful treatment of offsets and low-frequency drift.", "teacher": "Qwen/Qwen2.5-0.5B-Instruct", "teacher_revision": "7ae557604adf67be50417f59c2c2f167def9a775"} | |
| {"id": "topic-13-style-0", "topic_id": 13, "split": "train", "prompt": "Explain train-validation-test splits and data leakage in three concise sentences for a junior engineer.", "source": "original Sofia prompt bank v1", "response": "Use training data to fit model parameters, validation data to choose settings, and a held-out test set for final reporting. Split related machines or topic families together to reduce leakage. Repeated test-driven tuning invalidates the intended hold-out comparison.", "teacher": "Qwen/Qwen2.5-0.5B-Instruct", "teacher_revision": "7ae557604adf67be50417f59c2c2f167def9a775"} | |
| {"id": "topic-13-style-1", "topic_id": 13, "split": "train", "prompt": "What is train-validation-test splits and data leakage? Give a short practical example.", "source": "original Sofia prompt bank v1", "response": "Use training data to fit model parameters, validation data to choose settings, and a held-out test set for final reporting. Split related machines or topic families together to reduce leakage. Repeated test-driven tuning invalidates the intended hold-out comparison.", "teacher": "Qwen/Qwen2.5-0.5B-Instruct", "teacher_revision": "7ae557604adf67be50417f59c2c2f167def9a775"} | |
| {"id": "topic-13-style-2", "topic_id": 13, "split": "train", "prompt": "Describe one common mistake involving train-validation-test splits and data leakage and how to avoid it.", "source": "original Sofia prompt bank v1", "response": "Use training data to fit model parameters, validation data to choose settings, and a held-out test set for final reporting. Split related machines or topic families together to reduce leakage. Repeated test-driven tuning invalidates the intended hold-out comparison.", "teacher": "Qwen/Qwen2.5-0.5B-Instruct", "teacher_revision": "7ae557604adf67be50417f59c2c2f167def9a775"} | |
| {"id": "topic-13-style-3", "topic_id": 13, "split": "train", "prompt": "Give a brief technical explanation of train-validation-test splits and data leakage. State an important limitation.", "source": "original Sofia prompt bank v1", "response": "Use training data to fit model parameters, validation data to choose settings, and a held-out test set for final reporting. Split related machines or topic families together to reduce leakage. Repeated test-driven tuning invalidates the intended hold-out comparison.", "teacher": "Qwen/Qwen2.5-0.5B-Instruct", "teacher_revision": "7ae557604adf67be50417f59c2c2f167def9a775"} | |
| {"id": "topic-14-style-0", "topic_id": 14, "split": "train", "prompt": "Explain normalizing features with training-set statistics only in three concise sentences for a junior engineer.", "source": "original Sofia prompt bank v1", "response": "Fit feature means and scales using the training partition only. Apply those same statistics to validation, test and deployment data. Computing the scaler from all observations leaks information about the evaluation distribution.", "teacher": "Qwen/Qwen2.5-0.5B-Instruct", "teacher_revision": "7ae557604adf67be50417f59c2c2f167def9a775"} | |
| {"id": "topic-14-style-1", "topic_id": 14, "split": "train", "prompt": "What is normalizing features with training-set statistics only? Give a short practical example.", "source": "original Sofia prompt bank v1", "response": "Fit feature means and scales using the training partition only. Apply those same statistics to validation, test and deployment data. Computing the scaler from all observations leaks information about the evaluation distribution.", "teacher": "Qwen/Qwen2.5-0.5B-Instruct", "teacher_revision": "7ae557604adf67be50417f59c2c2f167def9a775"} | |
| {"id": "topic-14-style-2", "topic_id": 14, "split": "train", "prompt": "Describe one common mistake involving normalizing features with training-set statistics only and how to avoid it.", "source": "original Sofia prompt bank v1", "response": "Fit feature means and scales using the training partition only. Apply those same statistics to validation, test and deployment data. Computing the scaler from all observations leaks information about the evaluation distribution.", "teacher": "Qwen/Qwen2.5-0.5B-Instruct", "teacher_revision": "7ae557604adf67be50417f59c2c2f167def9a775"} | |
| {"id": "topic-14-style-3", "topic_id": 14, "split": "train", "prompt": "Give a brief technical explanation of normalizing features with training-set statistics only. State an important limitation.", "source": "original Sofia prompt bank v1", "response": "Fit feature means and scales using the training partition only. Apply those same statistics to validation, test and deployment data. Computing the scaler from all observations leaks information about the evaluation distribution.", "teacher": "Qwen/Qwen2.5-0.5B-Instruct", "teacher_revision": "7ae557604adf67be50417f59c2c2f167def9a775"} | |
| {"id": "topic-15-style-0", "topic_id": 15, "split": "train", "prompt": "Explain softmax probability versus calibrated confidence in three concise sentences for a junior engineer.", "source": "original Sofia prompt bank v1", "response": "Softmax maps logits to nonnegative values that sum to one. A high softmax score is not evidence of calibrated confidence or correctness on unfamiliar inputs. Check calibration and distribution shift using separate evaluation data.", "teacher": "Qwen/Qwen2.5-0.5B-Instruct", "teacher_revision": "7ae557604adf67be50417f59c2c2f167def9a775"} | |
| {"id": "topic-15-style-1", "topic_id": 15, "split": "train", "prompt": "What is softmax probability versus calibrated confidence? Give a short practical example.", "source": "original Sofia prompt bank v1", "response": "Softmax maps logits to nonnegative values that sum to one. A high softmax score is not evidence of calibrated confidence or correctness on unfamiliar inputs. Check calibration and distribution shift using separate evaluation data.", "teacher": "Qwen/Qwen2.5-0.5B-Instruct", "teacher_revision": "7ae557604adf67be50417f59c2c2f167def9a775"} | |
| {"id": "topic-15-style-2", "topic_id": 15, "split": "train", "prompt": "Describe one common mistake involving softmax probability versus calibrated confidence and how to avoid it.", "source": "original Sofia prompt bank v1", "response": "Softmax maps logits to nonnegative values that sum to one. A high softmax score is not evidence of calibrated confidence or correctness on unfamiliar inputs. Check calibration and distribution shift using separate evaluation data.", "teacher": "Qwen/Qwen2.5-0.5B-Instruct", "teacher_revision": "7ae557604adf67be50417f59c2c2f167def9a775"} | |
| {"id": "topic-15-style-3", "topic_id": 15, "split": "train", "prompt": "Give a brief technical explanation of softmax probability versus calibrated confidence. State an important limitation.", "source": "original Sofia prompt bank v1", "response": "Softmax maps logits to nonnegative values that sum to one. A high softmax score is not evidence of calibrated confidence or correctness on unfamiliar inputs. Check calibration and distribution shift using separate evaluation data.", "teacher": "Qwen/Qwen2.5-0.5B-Instruct", "teacher_revision": "7ae557604adf67be50417f59c2c2f167def9a775"} | |
| {"id": "topic-16-style-0", "topic_id": 16, "split": "train", "prompt": "Explain temperature scaling in knowledge distillation in three concise sentences for a junior engineer.", "source": "original Sofia prompt bank v1", "response": "Temperature T softens teacher probabilities by applying softmax(logits/T). Higher T exposes probability mass outside the top class. In standard logit distillation the KL term is commonly multiplied by T squared to adjust gradient scale.", "teacher": "Qwen/Qwen2.5-0.5B-Instruct", "teacher_revision": "7ae557604adf67be50417f59c2c2f167def9a775"} | |
| {"id": "topic-16-style-1", "topic_id": 16, "split": "train", "prompt": "What is temperature scaling in knowledge distillation? Give a short practical example.", "source": "original Sofia prompt bank v1", "response": "Temperature T softens teacher probabilities by applying softmax(logits/T). Higher T exposes probability mass outside the top class. In standard logit distillation the KL term is commonly multiplied by T squared to adjust gradient scale.", "teacher": "Qwen/Qwen2.5-0.5B-Instruct", "teacher_revision": "7ae557604adf67be50417f59c2c2f167def9a775"} | |
| {"id": "topic-16-style-2", "topic_id": 16, "split": "train", "prompt": "Describe one common mistake involving temperature scaling in knowledge distillation and how to avoid it.", "source": "original Sofia prompt bank v1", "response": "Temperature T softens teacher probabilities by applying softmax(logits/T). Higher T exposes probability mass outside the top class. In standard logit distillation the KL term is commonly multiplied by T squared to adjust gradient scale.", "teacher": "Qwen/Qwen2.5-0.5B-Instruct", "teacher_revision": "7ae557604adf67be50417f59c2c2f167def9a775"} | |
| {"id": "topic-16-style-3", "topic_id": 16, "split": "train", "prompt": "Give a brief technical explanation of temperature scaling in knowledge distillation. State an important limitation.", "source": "original Sofia prompt bank v1", "response": "Temperature T softens teacher probabilities by applying softmax(logits/T). Higher T exposes probability mass outside the top class. In standard logit distillation the KL term is commonly multiplied by T squared to adjust gradient scale.", "teacher": "Qwen/Qwen2.5-0.5B-Instruct", "teacher_revision": "7ae557604adf67be50417f59c2c2f167def9a775"} | |
| {"id": "topic-17-style-0", "topic_id": 17, "split": "train", "prompt": "Explain KL divergence between teacher and student probabilities in three concise sentences for a junior engineer.", "source": "original Sofia prompt bank v1", "response": "KL(p_teacher || p_student) is sum(p_teacher*log(p_teacher/p_student)). It is asymmetric and compares probability distributions, not hard labels alone. Mask prompt and padding positions when distilling assistant responses.", "teacher": "Qwen/Qwen2.5-0.5B-Instruct", "teacher_revision": "7ae557604adf67be50417f59c2c2f167def9a775"} | |
| {"id": "topic-17-style-1", "topic_id": 17, "split": "train", "prompt": "What is KL divergence between teacher and student probabilities? Give a short practical example.", "source": "original Sofia prompt bank v1", "response": "KL(p_teacher || p_student) is sum(p_teacher*log(p_teacher/p_student)). It is asymmetric and compares probability distributions, not hard labels alone. Mask prompt and padding positions when distilling assistant responses.", "teacher": "Qwen/Qwen2.5-0.5B-Instruct", "teacher_revision": "7ae557604adf67be50417f59c2c2f167def9a775"} | |
| {"id": "topic-17-style-2", "topic_id": 17, "split": "train", "prompt": "Describe one common mistake involving KL divergence between teacher and student probabilities and how to avoid it.", "source": "original Sofia prompt bank v1", "response": "KL(p_teacher || p_student) is sum(p_teacher*log(p_teacher/p_student)). It is asymmetric and compares probability distributions, not hard labels alone. Mask prompt and padding positions when distilling assistant responses.", "teacher": "Qwen/Qwen2.5-0.5B-Instruct", "teacher_revision": "7ae557604adf67be50417f59c2c2f167def9a775"} | |
| {"id": "topic-17-style-3", "topic_id": 17, "split": "train", "prompt": "Give a brief technical explanation of KL divergence between teacher and student probabilities. State an important limitation.", "source": "original Sofia prompt bank v1", "response": "KL(p_teacher || p_student) is sum(p_teacher*log(p_teacher/p_student)). It is asymmetric and compares probability distributions, not hard labels alone. Mask prompt and padding positions when distilling assistant responses.", "teacher": "Qwen/Qwen2.5-0.5B-Instruct", "teacher_revision": "7ae557604adf67be50417f59c2c2f167def9a775"} | |
| {"id": "topic-18-style-0", "topic_id": 18, "split": "train", "prompt": "Explain cross-entropy for multiclass classification in three concise sentences for a junior engineer.", "source": "original Sofia prompt bank v1", "response": "For one correct class y, cross-entropy is -log(p[y]). It penalizes confident incorrect predictions and is usually averaged over samples or valid tokens. Class imbalance and label quality affect its interpretation.", "teacher": "Qwen/Qwen2.5-0.5B-Instruct", "teacher_revision": "7ae557604adf67be50417f59c2c2f167def9a775"} | |
| {"id": "topic-18-style-1", "topic_id": 18, "split": "train", "prompt": "What is cross-entropy for multiclass classification? Give a short practical example.", "source": "original Sofia prompt bank v1", "response": "For one correct class y, cross-entropy is -log(p[y]). It penalizes confident incorrect predictions and is usually averaged over samples or valid tokens. Class imbalance and label quality affect its interpretation.", "teacher": "Qwen/Qwen2.5-0.5B-Instruct", "teacher_revision": "7ae557604adf67be50417f59c2c2f167def9a775"} | |
| {"id": "topic-18-style-2", "topic_id": 18, "split": "train", "prompt": "Describe one common mistake involving cross-entropy for multiclass classification and how to avoid it.", "source": "original Sofia prompt bank v1", "response": "For one correct class y, cross-entropy is -log(p[y]). It penalizes confident incorrect predictions and is usually averaged over samples or valid tokens. Class imbalance and label quality affect its interpretation.", "teacher": "Qwen/Qwen2.5-0.5B-Instruct", "teacher_revision": "7ae557604adf67be50417f59c2c2f167def9a775"} | |
| {"id": "topic-18-style-3", "topic_id": 18, "split": "train", "prompt": "Give a brief technical explanation of cross-entropy for multiclass classification. State an important limitation.", "source": "original Sofia prompt bank v1", "response": "For one correct class y, cross-entropy is -log(p[y]). It penalizes confident incorrect predictions and is usually averaged over samples or valid tokens. Class imbalance and label quality affect its interpretation.", "teacher": "Qwen/Qwen2.5-0.5B-Instruct", "teacher_revision": "7ae557604adf67be50417f59c2c2f167def9a775"} | |
| {"id": "topic-19-style-0", "topic_id": 19, "split": "test", "prompt": "Explain LoRA and low-rank parameter adaptation in three concise sentences for a junior engineer.", "source": "original Sofia prompt bank v1", "response": "LoRA trains low-rank updates to selected weight matrices while freezing the base model. This reduces trainable parameters and optimizer memory. A merged adapter changes the full checkpoint and still depends on the original pretrained weights.", "teacher": "Qwen/Qwen2.5-0.5B-Instruct", "teacher_revision": "7ae557604adf67be50417f59c2c2f167def9a775"} | |
| {"id": "topic-19-style-1", "topic_id": 19, "split": "test", "prompt": "What is LoRA and low-rank parameter adaptation? Give a short practical example.", "source": "original Sofia prompt bank v1", "response": "LoRA trains low-rank updates to selected weight matrices while freezing the base model. This reduces trainable parameters and optimizer memory. A merged adapter changes the full checkpoint and still depends on the original pretrained weights.", "teacher": "Qwen/Qwen2.5-0.5B-Instruct", "teacher_revision": "7ae557604adf67be50417f59c2c2f167def9a775"} | |
| {"id": "topic-19-style-2", "topic_id": 19, "split": "test", "prompt": "Describe one common mistake involving LoRA and low-rank parameter adaptation and how to avoid it.", "source": "original Sofia prompt bank v1", "response": "LoRA trains low-rank updates to selected weight matrices while freezing the base model. This reduces trainable parameters and optimizer memory. A merged adapter changes the full checkpoint and still depends on the original pretrained weights.", "teacher": "Qwen/Qwen2.5-0.5B-Instruct", "teacher_revision": "7ae557604adf67be50417f59c2c2f167def9a775"} | |
| {"id": "topic-19-style-3", "topic_id": 19, "split": "test", "prompt": "Give a brief technical explanation of LoRA and low-rank parameter adaptation. State an important limitation.", "source": "original Sofia prompt bank v1", "response": "LoRA trains low-rank updates to selected weight matrices while freezing the base model. This reduces trainable parameters and optimizer memory. A merged adapter changes the full checkpoint and still depends on the original pretrained weights.", "teacher": "Qwen/Qwen2.5-0.5B-Instruct", "teacher_revision": "7ae557604adf67be50417f59c2c2f167def9a775"} | |
| {"id": "topic-20-style-0", "topic_id": 20, "split": "train", "prompt": "Explain layer pruning in a transformer language model in three concise sentences for a junior engineer.", "source": "original Sofia prompt bank v1", "response": "Layer pruning removes transformer blocks to reduce parameter count and compute. Copied embeddings and surviving blocks retain some pretrained structure. Distillation can repair part of the damage, but broad quality must be measured rather than assumed.", "teacher": "Qwen/Qwen2.5-0.5B-Instruct", "teacher_revision": "7ae557604adf67be50417f59c2c2f167def9a775"} | |
| {"id": "topic-20-style-1", "topic_id": 20, "split": "train", "prompt": "What is layer pruning in a transformer language model? Give a short practical example.", "source": "original Sofia prompt bank v1", "response": "Layer pruning removes transformer blocks to reduce parameter count and compute. Copied embeddings and surviving blocks retain some pretrained structure. Distillation can repair part of the damage, but broad quality must be measured rather than assumed.", "teacher": "Qwen/Qwen2.5-0.5B-Instruct", "teacher_revision": "7ae557604adf67be50417f59c2c2f167def9a775"} | |
| {"id": "topic-20-style-2", "topic_id": 20, "split": "train", "prompt": "Describe one common mistake involving layer pruning in a transformer language model and how to avoid it.", "source": "original Sofia prompt bank v1", "response": "Layer pruning removes transformer blocks to reduce parameter count and compute. Copied embeddings and surviving blocks retain some pretrained structure. Distillation can repair part of the damage, but broad quality must be measured rather than assumed.", "teacher": "Qwen/Qwen2.5-0.5B-Instruct", "teacher_revision": "7ae557604adf67be50417f59c2c2f167def9a775"} | |
| {"id": "topic-20-style-3", "topic_id": 20, "split": "train", "prompt": "Give a brief technical explanation of layer pruning in a transformer language model. State an important limitation.", "source": "original Sofia prompt bank v1", "response": "Layer pruning removes transformer blocks to reduce parameter count and compute. Copied embeddings and surviving blocks retain some pretrained structure. Distillation can repair part of the damage, but broad quality must be measured rather than assumed.", "teacher": "Qwen/Qwen2.5-0.5B-Instruct", "teacher_revision": "7ae557604adf67be50417f59c2c2f167def9a775"} | |
| {"id": "topic-21-style-0", "topic_id": 21, "split": "train", "prompt": "Explain why a small distillation run is not foundation-model pretraining in three concise sentences for a junior engineer.", "source": "original Sofia prompt bank v1", "response": "A short distillation run adapts a student using a pretrained teacher or inherited weights. Foundation pretraining from random initialization requires a substantially broader corpus and compute budget. Report the teacher, initialization, data volume and measured scope explicitly.", "teacher": "Qwen/Qwen2.5-0.5B-Instruct", "teacher_revision": "7ae557604adf67be50417f59c2c2f167def9a775"} | |
| {"id": "topic-21-style-1", "topic_id": 21, "split": "train", "prompt": "What is why a small distillation run is not foundation-model pretraining? Give a short practical example.", "source": "original Sofia prompt bank v1", "response": "A short distillation run adapts a student using a pretrained teacher or inherited weights. Foundation pretraining from random initialization requires a substantially broader corpus and compute budget. Report the teacher, initialization, data volume and measured scope explicitly.", "teacher": "Qwen/Qwen2.5-0.5B-Instruct", "teacher_revision": "7ae557604adf67be50417f59c2c2f167def9a775"} | |
| {"id": "topic-21-style-2", "topic_id": 21, "split": "train", "prompt": "Describe one common mistake involving why a small distillation run is not foundation-model pretraining and how to avoid it.", "source": "original Sofia prompt bank v1", "response": "A short distillation run adapts a student using a pretrained teacher or inherited weights. Foundation pretraining from random initialization requires a substantially broader corpus and compute budget. Report the teacher, initialization, data volume and measured scope explicitly.", "teacher": "Qwen/Qwen2.5-0.5B-Instruct", "teacher_revision": "7ae557604adf67be50417f59c2c2f167def9a775"} | |
| {"id": "topic-21-style-3", "topic_id": 21, "split": "train", "prompt": "Give a brief technical explanation of why a small distillation run is not foundation-model pretraining. State an important limitation.", "source": "original Sofia prompt bank v1", "response": "A short distillation run adapts a student using a pretrained teacher or inherited weights. Foundation pretraining from random initialization requires a substantially broader corpus and compute budget. Report the teacher, initialization, data volume and measured scope explicitly.", "teacher": "Qwen/Qwen2.5-0.5B-Instruct", "teacher_revision": "7ae557604adf67be50417f59c2c2f167def9a775"} | |
| {"id": "topic-22-style-0", "topic_id": 22, "split": "train", "prompt": "Explain SHA-256 checksums for model artifacts in three concise sentences for a junior engineer.", "source": "original Sofia prompt bank v1", "response": "SHA-256 computes a digest of exact artifact bytes. Comparing a downloaded weight file with a trusted expected digest detects changes. A digest supplied by the same untrusted source does not independently authenticate authorship.", "teacher": "Qwen/Qwen2.5-0.5B-Instruct", "teacher_revision": "7ae557604adf67be50417f59c2c2f167def9a775"} | |
| {"id": "topic-22-style-1", "topic_id": 22, "split": "train", "prompt": "What is SHA-256 checksums for model artifacts? Give a short practical example.", "source": "original Sofia prompt bank v1", "response": "SHA-256 computes a digest of exact artifact bytes. Comparing a downloaded weight file with a trusted expected digest detects changes. A digest supplied by the same untrusted source does not independently authenticate authorship.", "teacher": "Qwen/Qwen2.5-0.5B-Instruct", "teacher_revision": "7ae557604adf67be50417f59c2c2f167def9a775"} | |
| {"id": "topic-22-style-2", "topic_id": 22, "split": "train", "prompt": "Describe one common mistake involving SHA-256 checksums for model artifacts and how to avoid it.", "source": "original Sofia prompt bank v1", "response": "SHA-256 computes a digest of exact artifact bytes. Comparing a downloaded weight file with a trusted expected digest detects changes. A digest supplied by the same untrusted source does not independently authenticate authorship.", "teacher": "Qwen/Qwen2.5-0.5B-Instruct", "teacher_revision": "7ae557604adf67be50417f59c2c2f167def9a775"} | |
| {"id": "topic-22-style-3", "topic_id": 22, "split": "train", "prompt": "Give a brief technical explanation of SHA-256 checksums for model artifacts. State an important limitation.", "source": "original Sofia prompt bank v1", "response": "SHA-256 computes a digest of exact artifact bytes. Comparing a downloaded weight file with a trusted expected digest detects changes. A digest supplied by the same untrusted source does not independently authenticate authorship.", "teacher": "Qwen/Qwen2.5-0.5B-Instruct", "teacher_revision": "7ae557604adf67be50417f59c2c2f167def9a775"} | |
| {"id": "topic-23-style-0", "topic_id": 23, "split": "train", "prompt": "Explain safetensors and the risks of loading untrusted pickle files in three concise sentences for a junior engineer.", "source": "original Sofia prompt bank v1", "response": "Safetensors stores tensor values without the general Python object deserialization used by pickle. Avoid loading untrusted pickle checkpoints because deserialization can execute code. File integrity and trusted provenance still need separate checks.", "teacher": "Qwen/Qwen2.5-0.5B-Instruct", "teacher_revision": "7ae557604adf67be50417f59c2c2f167def9a775"} | |
| {"id": "topic-23-style-1", "topic_id": 23, "split": "train", "prompt": "What is safetensors and the risks of loading untrusted pickle files? Give a short practical example.", "source": "original Sofia prompt bank v1", "response": "Safetensors stores tensor values without the general Python object deserialization used by pickle. Avoid loading untrusted pickle checkpoints because deserialization can execute code. File integrity and trusted provenance still need separate checks.", "teacher": "Qwen/Qwen2.5-0.5B-Instruct", "teacher_revision": "7ae557604adf67be50417f59c2c2f167def9a775"} | |
| {"id": "topic-23-style-2", "topic_id": 23, "split": "train", "prompt": "Describe one common mistake involving safetensors and the risks of loading untrusted pickle files and how to avoid it.", "source": "original Sofia prompt bank v1", "response": "Safetensors stores tensor values without the general Python object deserialization used by pickle. Avoid loading untrusted pickle checkpoints because deserialization can execute code. File integrity and trusted provenance still need separate checks.", "teacher": "Qwen/Qwen2.5-0.5B-Instruct", "teacher_revision": "7ae557604adf67be50417f59c2c2f167def9a775"} | |
| {"id": "topic-23-style-3", "topic_id": 23, "split": "train", "prompt": "Give a brief technical explanation of safetensors and the risks of loading untrusted pickle files. State an important limitation.", "source": "original Sofia prompt bank v1", "response": "Safetensors stores tensor values without the general Python object deserialization used by pickle. Avoid loading untrusted pickle checkpoints because deserialization can execute code. File integrity and trusted provenance still need separate checks.", "teacher": "Qwen/Qwen2.5-0.5B-Instruct", "teacher_revision": "7ae557604adf67be50417f59c2c2f167def9a775"} | |
| {"id": "topic-24-style-0", "topic_id": 24, "split": "train", "prompt": "Explain NumPy broadcasting for vectorized feature calculations in three concise sentences for a junior engineer.", "source": "original Sofia prompt bank v1", "response": "NumPy broadcasting aligns compatible array dimensions for vectorized operations. For X of shape (n,d), subtracting a mean vector of shape (d,) normalizes each row. Validate shapes because accidental broadcasting can silently compute the wrong result.", "teacher": "Qwen/Qwen2.5-0.5B-Instruct", "teacher_revision": "7ae557604adf67be50417f59c2c2f167def9a775"} | |
| {"id": "topic-24-style-1", "topic_id": 24, "split": "train", "prompt": "What is NumPy broadcasting for vectorized feature calculations? Give a short practical example.", "source": "original Sofia prompt bank v1", "response": "NumPy broadcasting aligns compatible array dimensions for vectorized operations. For X of shape (n,d), subtracting a mean vector of shape (d,) normalizes each row. Validate shapes because accidental broadcasting can silently compute the wrong result.", "teacher": "Qwen/Qwen2.5-0.5B-Instruct", "teacher_revision": "7ae557604adf67be50417f59c2c2f167def9a775"} | |
| {"id": "topic-24-style-2", "topic_id": 24, "split": "train", "prompt": "Describe one common mistake involving NumPy broadcasting for vectorized feature calculations and how to avoid it.", "source": "original Sofia prompt bank v1", "response": "NumPy broadcasting aligns compatible array dimensions for vectorized operations. For X of shape (n,d), subtracting a mean vector of shape (d,) normalizes each row. Validate shapes because accidental broadcasting can silently compute the wrong result.", "teacher": "Qwen/Qwen2.5-0.5B-Instruct", "teacher_revision": "7ae557604adf67be50417f59c2c2f167def9a775"} | |
| {"id": "topic-24-style-3", "topic_id": 24, "split": "train", "prompt": "Give a brief technical explanation of NumPy broadcasting for vectorized feature calculations. State an important limitation.", "source": "original Sofia prompt bank v1", "response": "NumPy broadcasting aligns compatible array dimensions for vectorized operations. For X of shape (n,d), subtracting a mean vector of shape (d,) normalizes each row. Validate shapes because accidental broadcasting can silently compute the wrong result.", "teacher": "Qwen/Qwen2.5-0.5B-Instruct", "teacher_revision": "7ae557604adf67be50417f59c2c2f167def9a775"} | |
| {"id": "topic-25-style-0", "topic_id": 25, "split": "validation", "prompt": "Explain finite-value validation in sensor pipelines in three concise sentences for a junior engineer.", "source": "original Sofia prompt bank v1", "response": "Reject NaN and infinity before numerical sensor calculations. Use np.isfinite(values).all() together with shape and unit checks. Replacing invalid measurements with arbitrary zeros can hide acquisition failures.", "teacher": "Qwen/Qwen2.5-0.5B-Instruct", "teacher_revision": "7ae557604adf67be50417f59c2c2f167def9a775"} | |
| {"id": "topic-25-style-1", "topic_id": 25, "split": "validation", "prompt": "What is finite-value validation in sensor pipelines? Give a short practical example.", "source": "original Sofia prompt bank v1", "response": "Reject NaN and infinity before numerical sensor calculations. Use np.isfinite(values).all() together with shape and unit checks. Replacing invalid measurements with arbitrary zeros can hide acquisition failures.", "teacher": "Qwen/Qwen2.5-0.5B-Instruct", "teacher_revision": "7ae557604adf67be50417f59c2c2f167def9a775"} | |
| {"id": "topic-25-style-2", "topic_id": 25, "split": "validation", "prompt": "Describe one common mistake involving finite-value validation in sensor pipelines and how to avoid it.", "source": "original Sofia prompt bank v1", "response": "Reject NaN and infinity before numerical sensor calculations. Use np.isfinite(values).all() together with shape and unit checks. Replacing invalid measurements with arbitrary zeros can hide acquisition failures.", "teacher": "Qwen/Qwen2.5-0.5B-Instruct", "teacher_revision": "7ae557604adf67be50417f59c2c2f167def9a775"} | |
| {"id": "topic-25-style-3", "topic_id": 25, "split": "validation", "prompt": "Give a brief technical explanation of finite-value validation in sensor pipelines. State an important limitation.", "source": "original Sofia prompt bank v1", "response": "Reject NaN and infinity before numerical sensor calculations. Use np.isfinite(values).all() together with shape and unit checks. Replacing invalid measurements with arbitrary zeros can hide acquisition failures.", "teacher": "Qwen/Qwen2.5-0.5B-Instruct", "teacher_revision": "7ae557604adf67be50417f59c2c2f167def9a775"} | |
| {"id": "topic-26-style-0", "topic_id": 26, "split": "train", "prompt": "Explain Python context managers for reading files in three concise sentences for a junior engineer.", "source": "original Sofia prompt bank v1", "response": "A Python with block calls a context manager's cleanup when execution leaves the block. Use with open(path) as f to close a file even if reading raises an exception. It does not make the file contents trustworthy.", "teacher": "Qwen/Qwen2.5-0.5B-Instruct", "teacher_revision": "7ae557604adf67be50417f59c2c2f167def9a775"} | |
| {"id": "topic-26-style-1", "topic_id": 26, "split": "train", "prompt": "What is Python context managers for reading files? Give a short practical example.", "source": "original Sofia prompt bank v1", "response": "A Python with block calls a context manager's cleanup when execution leaves the block. Use with open(path) as f to close a file even if reading raises an exception. It does not make the file contents trustworthy.", "teacher": "Qwen/Qwen2.5-0.5B-Instruct", "teacher_revision": "7ae557604adf67be50417f59c2c2f167def9a775"} | |
| {"id": "topic-26-style-2", "topic_id": 26, "split": "train", "prompt": "Describe one common mistake involving Python context managers for reading files and how to avoid it.", "source": "original Sofia prompt bank v1", "response": "A Python with block calls a context manager's cleanup when execution leaves the block. Use with open(path) as f to close a file even if reading raises an exception. It does not make the file contents trustworthy.", "teacher": "Qwen/Qwen2.5-0.5B-Instruct", "teacher_revision": "7ae557604adf67be50417f59c2c2f167def9a775"} | |
| {"id": "topic-26-style-3", "topic_id": 26, "split": "train", "prompt": "Give a brief technical explanation of Python context managers for reading files. State an important limitation.", "source": "original Sofia prompt bank v1", "response": "A Python with block calls a context manager's cleanup when execution leaves the block. Use with open(path) as f to close a file even if reading raises an exception. It does not make the file contents trustworthy.", "teacher": "Qwen/Qwen2.5-0.5B-Instruct", "teacher_revision": "7ae557604adf67be50417f59c2c2f167def9a775"} | |
| {"id": "topic-27-style-0", "topic_id": 27, "split": "test", "prompt": "Explain the difference between a list and a NumPy array in three concise sentences for a junior engineer.", "source": "original Sofia prompt bank v1", "response": "A Python list is a general sequence of object references; a NumPy array is a typed numerical container. NumPy arrays support vectorized arithmetic and explicit shapes. Multiplying a list repeats it, while multiplying a numeric array scales its elements.", "teacher": "Qwen/Qwen2.5-0.5B-Instruct", "teacher_revision": "7ae557604adf67be50417f59c2c2f167def9a775"} | |
| {"id": "topic-27-style-1", "topic_id": 27, "split": "test", "prompt": "What is the difference between a list and a NumPy array? Give a short practical example.", "source": "original Sofia prompt bank v1", "response": "A Python list is a general sequence of object references; a NumPy array is a typed numerical container. NumPy arrays support vectorized arithmetic and explicit shapes. Multiplying a list repeats it, while multiplying a numeric array scales its elements.", "teacher": "Qwen/Qwen2.5-0.5B-Instruct", "teacher_revision": "7ae557604adf67be50417f59c2c2f167def9a775"} | |
| {"id": "topic-27-style-2", "topic_id": 27, "split": "test", "prompt": "Describe one common mistake involving the difference between a list and a NumPy array and how to avoid it.", "source": "original Sofia prompt bank v1", "response": "A Python list is a general sequence of object references; a NumPy array is a typed numerical container. NumPy arrays support vectorized arithmetic and explicit shapes. Multiplying a list repeats it, while multiplying a numeric array scales its elements.", "teacher": "Qwen/Qwen2.5-0.5B-Instruct", "teacher_revision": "7ae557604adf67be50417f59c2c2f167def9a775"} | |
| {"id": "topic-27-style-3", "topic_id": 27, "split": "test", "prompt": "Give a brief technical explanation of the difference between a list and a NumPy array. State an important limitation.", "source": "original Sofia prompt bank v1", "response": "A Python list is a general sequence of object references; a NumPy array is a typed numerical container. NumPy arrays support vectorized arithmetic and explicit shapes. Multiplying a list repeats it, while multiplying a numeric array scales its elements.", "teacher": "Qwen/Qwen2.5-0.5B-Instruct", "teacher_revision": "7ae557604adf67be50417f59c2c2f167def9a775"} | |
| {"id": "topic-28-style-0", "topic_id": 28, "split": "train", "prompt": "Explain unit tests for numerical algorithms in three concise sentences for a junior engineer.", "source": "original Sofia prompt bank v1", "response": "Numerical tests should compare against independent formulas, analytical special cases and meaningful tolerances. Include invalid inputs and invariants such as energy or symmetry. A test that repeats the implementation's own logic may reproduce the same mistake.", "teacher": "Qwen/Qwen2.5-0.5B-Instruct", "teacher_revision": "7ae557604adf67be50417f59c2c2f167def9a775"} | |
| {"id": "topic-28-style-1", "topic_id": 28, "split": "train", "prompt": "What is unit tests for numerical algorithms? Give a short practical example.", "source": "original Sofia prompt bank v1", "response": "Numerical tests should compare against independent formulas, analytical special cases and meaningful tolerances. Include invalid inputs and invariants such as energy or symmetry. A test that repeats the implementation's own logic may reproduce the same mistake.", "teacher": "Qwen/Qwen2.5-0.5B-Instruct", "teacher_revision": "7ae557604adf67be50417f59c2c2f167def9a775"} | |
| {"id": "topic-28-style-2", "topic_id": 28, "split": "train", "prompt": "Describe one common mistake involving unit tests for numerical algorithms and how to avoid it.", "source": "original Sofia prompt bank v1", "response": "Numerical tests should compare against independent formulas, analytical special cases and meaningful tolerances. Include invalid inputs and invariants such as energy or symmetry. A test that repeats the implementation's own logic may reproduce the same mistake.", "teacher": "Qwen/Qwen2.5-0.5B-Instruct", "teacher_revision": "7ae557604adf67be50417f59c2c2f167def9a775"} | |
| {"id": "topic-28-style-3", "topic_id": 28, "split": "train", "prompt": "Give a brief technical explanation of unit tests for numerical algorithms. State an important limitation.", "source": "original Sofia prompt bank v1", "response": "Numerical tests should compare against independent formulas, analytical special cases and meaningful tolerances. Include invalid inputs and invariants such as energy or symmetry. A test that repeats the implementation's own logic may reproduce the same mistake.", "teacher": "Qwen/Qwen2.5-0.5B-Instruct", "teacher_revision": "7ae557604adf67be50417f59c2c2f167def9a775"} | |
| {"id": "topic-29-style-0", "topic_id": 29, "split": "train", "prompt": "Explain JSON serialization of floating-point measurements in three concise sentences for a junior engineer.", "source": "original Sofia prompt bank v1", "response": "JSON transports structured data but does not encode physical units automatically. Serialize finite measurements with explicit unit, timestamp and quality fields. Reject nonstandard NaN or infinity values when strict interoperable JSON is required.", "teacher": "Qwen/Qwen2.5-0.5B-Instruct", "teacher_revision": "7ae557604adf67be50417f59c2c2f167def9a775"} | |
| {"id": "topic-29-style-1", "topic_id": 29, "split": "train", "prompt": "What is JSON serialization of floating-point measurements? Give a short practical example.", "source": "original Sofia prompt bank v1", "response": "JSON transports structured data but does not encode physical units automatically. Serialize finite measurements with explicit unit, timestamp and quality fields. Reject nonstandard NaN or infinity values when strict interoperable JSON is required.", "teacher": "Qwen/Qwen2.5-0.5B-Instruct", "teacher_revision": "7ae557604adf67be50417f59c2c2f167def9a775"} | |
| {"id": "topic-29-style-2", "topic_id": 29, "split": "train", "prompt": "Describe one common mistake involving JSON serialization of floating-point measurements and how to avoid it.", "source": "original Sofia prompt bank v1", "response": "JSON transports structured data but does not encode physical units automatically. Serialize finite measurements with explicit unit, timestamp and quality fields. Reject nonstandard NaN or infinity values when strict interoperable JSON is required.", "teacher": "Qwen/Qwen2.5-0.5B-Instruct", "teacher_revision": "7ae557604adf67be50417f59c2c2f167def9a775"} | |
| {"id": "topic-29-style-3", "topic_id": 29, "split": "train", "prompt": "Give a brief technical explanation of JSON serialization of floating-point measurements. State an important limitation.", "source": "original Sofia prompt bank v1", "response": "JSON transports structured data but does not encode physical units automatically. Serialize finite measurements with explicit unit, timestamp and quality fields. Reject nonstandard NaN or infinity values when strict interoperable JSON is required.", "teacher": "Qwen/Qwen2.5-0.5B-Instruct", "teacher_revision": "7ae557604adf67be50417f59c2c2f167def9a775"} | |
| {"id": "topic-30-style-0", "topic_id": 30, "split": "train", "prompt": "Explain MQTT quality-of-service levels in three concise sentences for a junior engineer.", "source": "original Sofia prompt bank v1", "response": "MQTT QoS 0 is at most once, QoS 1 is at least once, and QoS 2 is exactly once within the protocol exchange. QoS 1 can deliver duplicates. Application-level persistence and idempotency remain important for end-to-end processing.", "teacher": "Qwen/Qwen2.5-0.5B-Instruct", "teacher_revision": "7ae557604adf67be50417f59c2c2f167def9a775"} | |
| {"id": "topic-30-style-1", "topic_id": 30, "split": "train", "prompt": "What is MQTT quality-of-service levels? Give a short practical example.", "source": "original Sofia prompt bank v1", "response": "MQTT QoS 0 is at most once, QoS 1 is at least once, and QoS 2 is exactly once within the protocol exchange. QoS 1 can deliver duplicates. Application-level persistence and idempotency remain important for end-to-end processing.", "teacher": "Qwen/Qwen2.5-0.5B-Instruct", "teacher_revision": "7ae557604adf67be50417f59c2c2f167def9a775"} | |
| {"id": "topic-30-style-2", "topic_id": 30, "split": "train", "prompt": "Describe one common mistake involving MQTT quality-of-service levels and how to avoid it.", "source": "original Sofia prompt bank v1", "response": "MQTT QoS 0 is at most once, QoS 1 is at least once, and QoS 2 is exactly once within the protocol exchange. QoS 1 can deliver duplicates. Application-level persistence and idempotency remain important for end-to-end processing.", "teacher": "Qwen/Qwen2.5-0.5B-Instruct", "teacher_revision": "7ae557604adf67be50417f59c2c2f167def9a775"} | |
| {"id": "topic-30-style-3", "topic_id": 30, "split": "train", "prompt": "Give a brief technical explanation of MQTT quality-of-service levels. State an important limitation.", "source": "original Sofia prompt bank v1", "response": "MQTT QoS 0 is at most once, QoS 1 is at least once, and QoS 2 is exactly once within the protocol exchange. QoS 1 can deliver duplicates. Application-level persistence and idempotency remain important for end-to-end processing.", "teacher": "Qwen/Qwen2.5-0.5B-Instruct", "teacher_revision": "7ae557604adf67be50417f59c2c2f167def9a775"} | |
| {"id": "topic-31-style-0", "topic_id": 31, "split": "train", "prompt": "Explain Modbus registers and endianness in three concise sentences for a junior engineer.", "source": "original Sofia prompt bank v1", "response": "A Modbus register contains 16 bits. A multiregister value needs a specified register order, byte interpretation and data type from the device map. Guessing those conventions can yield plausible but incorrect numerical values.", "teacher": "Qwen/Qwen2.5-0.5B-Instruct", "teacher_revision": "7ae557604adf67be50417f59c2c2f167def9a775"} | |
| {"id": "topic-31-style-1", "topic_id": 31, "split": "train", "prompt": "What is Modbus registers and endianness? Give a short practical example.", "source": "original Sofia prompt bank v1", "response": "A Modbus register contains 16 bits. A multiregister value needs a specified register order, byte interpretation and data type from the device map. Guessing those conventions can yield plausible but incorrect numerical values.", "teacher": "Qwen/Qwen2.5-0.5B-Instruct", "teacher_revision": "7ae557604adf67be50417f59c2c2f167def9a775"} | |
| {"id": "topic-31-style-2", "topic_id": 31, "split": "train", "prompt": "Describe one common mistake involving Modbus registers and endianness and how to avoid it.", "source": "original Sofia prompt bank v1", "response": "A Modbus register contains 16 bits. A multiregister value needs a specified register order, byte interpretation and data type from the device map. Guessing those conventions can yield plausible but incorrect numerical values.", "teacher": "Qwen/Qwen2.5-0.5B-Instruct", "teacher_revision": "7ae557604adf67be50417f59c2c2f167def9a775"} | |
| {"id": "topic-31-style-3", "topic_id": 31, "split": "train", "prompt": "Give a brief technical explanation of Modbus registers and endianness. State an important limitation.", "source": "original Sofia prompt bank v1", "response": "A Modbus register contains 16 bits. A multiregister value needs a specified register order, byte interpretation and data type from the device map. Guessing those conventions can yield plausible but incorrect numerical values.", "teacher": "Qwen/Qwen2.5-0.5B-Instruct", "teacher_revision": "7ae557604adf67be50417f59c2c2f167def9a775"} | |
| {"id": "topic-32-style-0", "topic_id": 32, "split": "train", "prompt": "Explain bounded buffers in edge telemetry systems in three concise sentences for a junior engineer.", "source": "original Sofia prompt bank v1", "response": "A bounded buffer limits the number of retained records. Choose and document overflow behavior such as drop-oldest, reject-newest or backpressure. Monitor dropped records because a memory bound alone does not preserve every measurement.", "teacher": "Qwen/Qwen2.5-0.5B-Instruct", "teacher_revision": "7ae557604adf67be50417f59c2c2f167def9a775"} | |
| {"id": "topic-32-style-1", "topic_id": 32, "split": "train", "prompt": "What is bounded buffers in edge telemetry systems? Give a short practical example.", "source": "original Sofia prompt bank v1", "response": "A bounded buffer limits the number of retained records. Choose and document overflow behavior such as drop-oldest, reject-newest or backpressure. Monitor dropped records because a memory bound alone does not preserve every measurement.", "teacher": "Qwen/Qwen2.5-0.5B-Instruct", "teacher_revision": "7ae557604adf67be50417f59c2c2f167def9a775"} | |
| {"id": "topic-32-style-2", "topic_id": 32, "split": "train", "prompt": "Describe one common mistake involving bounded buffers in edge telemetry systems and how to avoid it.", "source": "original Sofia prompt bank v1", "response": "A bounded buffer limits the number of retained records. Choose and document overflow behavior such as drop-oldest, reject-newest or backpressure. Monitor dropped records because a memory bound alone does not preserve every measurement.", "teacher": "Qwen/Qwen2.5-0.5B-Instruct", "teacher_revision": "7ae557604adf67be50417f59c2c2f167def9a775"} | |
| {"id": "topic-32-style-3", "topic_id": 32, "split": "train", "prompt": "Give a brief technical explanation of bounded buffers in edge telemetry systems. State an important limitation.", "source": "original Sofia prompt bank v1", "response": "A bounded buffer limits the number of retained records. Choose and document overflow behavior such as drop-oldest, reject-newest or backpressure. Monitor dropped records because a memory bound alone does not preserve every measurement.", "teacher": "Qwen/Qwen2.5-0.5B-Instruct", "teacher_revision": "7ae557604adf67be50417f59c2c2f167def9a775"} | |
| {"id": "topic-33-style-0", "topic_id": 33, "split": "train", "prompt": "Explain monotonic clocks versus wall clocks in three concise sentences for a junior engineer.", "source": "original Sofia prompt bank v1", "response": "A wall clock represents calendar time and can jump after synchronization. A monotonic clock measures elapsed intervals without going backward. Use monotonic time for local deadlines and record wall timestamps separately for event correlation.", "teacher": "Qwen/Qwen2.5-0.5B-Instruct", "teacher_revision": "7ae557604adf67be50417f59c2c2f167def9a775"} | |
| {"id": "topic-33-style-1", "topic_id": 33, "split": "train", "prompt": "What is monotonic clocks versus wall clocks? Give a short practical example.", "source": "original Sofia prompt bank v1", "response": "A wall clock represents calendar time and can jump after synchronization. A monotonic clock measures elapsed intervals without going backward. Use monotonic time for local deadlines and record wall timestamps separately for event correlation.", "teacher": "Qwen/Qwen2.5-0.5B-Instruct", "teacher_revision": "7ae557604adf67be50417f59c2c2f167def9a775"} | |
| {"id": "topic-33-style-2", "topic_id": 33, "split": "train", "prompt": "Describe one common mistake involving monotonic clocks versus wall clocks and how to avoid it.", "source": "original Sofia prompt bank v1", "response": "A wall clock represents calendar time and can jump after synchronization. A monotonic clock measures elapsed intervals without going backward. Use monotonic time for local deadlines and record wall timestamps separately for event correlation.", "teacher": "Qwen/Qwen2.5-0.5B-Instruct", "teacher_revision": "7ae557604adf67be50417f59c2c2f167def9a775"} | |
| {"id": "topic-33-style-3", "topic_id": 33, "split": "train", "prompt": "Give a brief technical explanation of monotonic clocks versus wall clocks. State an important limitation.", "source": "original Sofia prompt bank v1", "response": "A wall clock represents calendar time and can jump after synchronization. A monotonic clock measures elapsed intervals without going backward. Use monotonic time for local deadlines and record wall timestamps separately for event correlation.", "teacher": "Qwen/Qwen2.5-0.5B-Instruct", "teacher_revision": "7ae557604adf67be50417f59c2c2f167def9a775"} | |
| {"id": "topic-34-style-0", "topic_id": 34, "split": "train", "prompt": "Explain why a language model should not directly actuate machinery in three concise sentences for a junior engineer.", "source": "original Sofia prompt bank v1", "response": "A language model can produce incorrect or unauthorized control instructions. Keep advisory text separate from a deterministic control interface with explicit authorization and physical interlocks. Model confidence must not bypass those controls.", "teacher": "Qwen/Qwen2.5-0.5B-Instruct", "teacher_revision": "7ae557604adf67be50417f59c2c2f167def9a775"} | |
| {"id": "topic-34-style-1", "topic_id": 34, "split": "train", "prompt": "What is why a language model should not directly actuate machinery? Give a short practical example.", "source": "original Sofia prompt bank v1", "response": "A language model can produce incorrect or unauthorized control instructions. Keep advisory text separate from a deterministic control interface with explicit authorization and physical interlocks. Model confidence must not bypass those controls.", "teacher": "Qwen/Qwen2.5-0.5B-Instruct", "teacher_revision": "7ae557604adf67be50417f59c2c2f167def9a775"} | |
| {"id": "topic-34-style-2", "topic_id": 34, "split": "train", "prompt": "Describe one common mistake involving why a language model should not directly actuate machinery and how to avoid it.", "source": "original Sofia prompt bank v1", "response": "A language model can produce incorrect or unauthorized control instructions. Keep advisory text separate from a deterministic control interface with explicit authorization and physical interlocks. Model confidence must not bypass those controls.", "teacher": "Qwen/Qwen2.5-0.5B-Instruct", "teacher_revision": "7ae557604adf67be50417f59c2c2f167def9a775"} | |
| {"id": "topic-34-style-3", "topic_id": 34, "split": "train", "prompt": "Give a brief technical explanation of why a language model should not directly actuate machinery. State an important limitation.", "source": "original Sofia prompt bank v1", "response": "A language model can produce incorrect or unauthorized control instructions. Keep advisory text separate from a deterministic control interface with explicit authorization and physical interlocks. Model confidence must not bypass those controls.", "teacher": "Qwen/Qwen2.5-0.5B-Instruct", "teacher_revision": "7ae557604adf67be50417f59c2c2f167def9a775"} | |
| {"id": "topic-35-style-0", "topic_id": 35, "split": "test", "prompt": "Explain default-deny authorization for control commands in three concise sentences for a junior engineer.", "source": "original Sofia prompt bank v1", "response": "Default-deny authorization rejects commands unless every required permission and condition is satisfied. Check operator authorization, command allowlists, freshness and interlocks before execution. Missing context must not silently become approval.", "teacher": "Qwen/Qwen2.5-0.5B-Instruct", "teacher_revision": "7ae557604adf67be50417f59c2c2f167def9a775"} | |
| {"id": "topic-35-style-1", "topic_id": 35, "split": "test", "prompt": "What is default-deny authorization for control commands? Give a short practical example.", "source": "original Sofia prompt bank v1", "response": "Default-deny authorization rejects commands unless every required permission and condition is satisfied. Check operator authorization, command allowlists, freshness and interlocks before execution. Missing context must not silently become approval.", "teacher": "Qwen/Qwen2.5-0.5B-Instruct", "teacher_revision": "7ae557604adf67be50417f59c2c2f167def9a775"} | |
| {"id": "topic-35-style-2", "topic_id": 35, "split": "test", "prompt": "Describe one common mistake involving default-deny authorization for control commands and how to avoid it.", "source": "original Sofia prompt bank v1", "response": "Default-deny authorization rejects commands unless every required permission and condition is satisfied. Check operator authorization, command allowlists, freshness and interlocks before execution. Missing context must not silently become approval.", "teacher": "Qwen/Qwen2.5-0.5B-Instruct", "teacher_revision": "7ae557604adf67be50417f59c2c2f167def9a775"} | |
| {"id": "topic-35-style-3", "topic_id": 35, "split": "test", "prompt": "Give a brief technical explanation of default-deny authorization for control commands. State an important limitation.", "source": "original Sofia prompt bank v1", "response": "Default-deny authorization rejects commands unless every required permission and condition is satisfied. Check operator authorization, command allowlists, freshness and interlocks before execution. Missing context must not silently become approval.", "teacher": "Qwen/Qwen2.5-0.5B-Instruct", "teacher_revision": "7ae557604adf67be50417f59c2c2f167def9a775"} | |
| {"id": "topic-36-style-0", "topic_id": 36, "split": "train", "prompt": "Explain harmonic distortion in electrical measurements in three concise sentences for a junior engineer.", "source": "original Sofia prompt bank v1", "response": "Total harmonic distortion compares the RMS contribution of higher harmonics with the fundamental RMS. State which harmonics and acquisition bandwidth are included. A distorted waveform needs adequate sampling and a well-defined fundamental estimate.", "teacher": "Qwen/Qwen2.5-0.5B-Instruct", "teacher_revision": "7ae557604adf67be50417f59c2c2f167def9a775"} | |
| {"id": "topic-36-style-1", "topic_id": 36, "split": "train", "prompt": "What is harmonic distortion in electrical measurements? Give a short practical example.", "source": "original Sofia prompt bank v1", "response": "Total harmonic distortion compares the RMS contribution of higher harmonics with the fundamental RMS. State which harmonics and acquisition bandwidth are included. A distorted waveform needs adequate sampling and a well-defined fundamental estimate.", "teacher": "Qwen/Qwen2.5-0.5B-Instruct", "teacher_revision": "7ae557604adf67be50417f59c2c2f167def9a775"} | |
| {"id": "topic-36-style-2", "topic_id": 36, "split": "train", "prompt": "Describe one common mistake involving harmonic distortion in electrical measurements and how to avoid it.", "source": "original Sofia prompt bank v1", "response": "Total harmonic distortion compares the RMS contribution of higher harmonics with the fundamental RMS. State which harmonics and acquisition bandwidth are included. A distorted waveform needs adequate sampling and a well-defined fundamental estimate.", "teacher": "Qwen/Qwen2.5-0.5B-Instruct", "teacher_revision": "7ae557604adf67be50417f59c2c2f167def9a775"} | |
| {"id": "topic-36-style-3", "topic_id": 36, "split": "train", "prompt": "Give a brief technical explanation of harmonic distortion in electrical measurements. State an important limitation.", "source": "original Sofia prompt bank v1", "response": "Total harmonic distortion compares the RMS contribution of higher harmonics with the fundamental RMS. State which harmonics and acquisition bandwidth are included. A distorted waveform needs adequate sampling and a well-defined fundamental estimate.", "teacher": "Qwen/Qwen2.5-0.5B-Instruct", "teacher_revision": "7ae557604adf67be50417f59c2c2f167def9a775"} | |
| {"id": "topic-37-style-0", "topic_id": 37, "split": "train", "prompt": "Explain power factor and active versus apparent power in three concise sentences for a junior engineer.", "source": "original Sofia prompt bank v1", "response": "Active power is the average instantaneous voltage-current product. Apparent power is Vrms*Irms, and true power factor is active power divided by apparent power. With waveform distortion, true power factor is not generally just the cosine of a phase angle.", "teacher": "Qwen/Qwen2.5-0.5B-Instruct", "teacher_revision": "7ae557604adf67be50417f59c2c2f167def9a775"} | |
| {"id": "topic-37-style-1", "topic_id": 37, "split": "train", "prompt": "What is power factor and active versus apparent power? Give a short practical example.", "source": "original Sofia prompt bank v1", "response": "Active power is the average instantaneous voltage-current product. Apparent power is Vrms*Irms, and true power factor is active power divided by apparent power. With waveform distortion, true power factor is not generally just the cosine of a phase angle.", "teacher": "Qwen/Qwen2.5-0.5B-Instruct", "teacher_revision": "7ae557604adf67be50417f59c2c2f167def9a775"} | |
| {"id": "topic-37-style-2", "topic_id": 37, "split": "train", "prompt": "Describe one common mistake involving power factor and active versus apparent power and how to avoid it.", "source": "original Sofia prompt bank v1", "response": "Active power is the average instantaneous voltage-current product. Apparent power is Vrms*Irms, and true power factor is active power divided by apparent power. With waveform distortion, true power factor is not generally just the cosine of a phase angle.", "teacher": "Qwen/Qwen2.5-0.5B-Instruct", "teacher_revision": "7ae557604adf67be50417f59c2c2f167def9a775"} | |
| {"id": "topic-37-style-3", "topic_id": 37, "split": "train", "prompt": "Give a brief technical explanation of power factor and active versus apparent power. State an important limitation.", "source": "original Sofia prompt bank v1", "response": "Active power is the average instantaneous voltage-current product. Apparent power is Vrms*Irms, and true power factor is active power divided by apparent power. With waveform distortion, true power factor is not generally just the cosine of a phase angle.", "teacher": "Qwen/Qwen2.5-0.5B-Instruct", "teacher_revision": "7ae557604adf67be50417f59c2c2f167def9a775"} | |
| {"id": "topic-38-style-0", "topic_id": 38, "split": "validation", "prompt": "Explain three-phase voltage imbalance in three concise sentences for a junior engineer.", "source": "original Sofia prompt bank v1", "response": "Three-phase voltage imbalance can be studied using positive- and negative-sequence components. A common ratio uses negative-sequence magnitude divided by positive-sequence magnitude. Specify the measurement definition rather than treating every imbalance metric as equivalent.", "teacher": "Qwen/Qwen2.5-0.5B-Instruct", "teacher_revision": "7ae557604adf67be50417f59c2c2f167def9a775"} | |
| {"id": "topic-38-style-1", "topic_id": 38, "split": "validation", "prompt": "What is three-phase voltage imbalance? Give a short practical example.", "source": "original Sofia prompt bank v1", "response": "Three-phase voltage imbalance can be studied using positive- and negative-sequence components. A common ratio uses negative-sequence magnitude divided by positive-sequence magnitude. Specify the measurement definition rather than treating every imbalance metric as equivalent.", "teacher": "Qwen/Qwen2.5-0.5B-Instruct", "teacher_revision": "7ae557604adf67be50417f59c2c2f167def9a775"} | |
| {"id": "topic-38-style-2", "topic_id": 38, "split": "validation", "prompt": "Describe one common mistake involving three-phase voltage imbalance and how to avoid it.", "source": "original Sofia prompt bank v1", "response": "Three-phase voltage imbalance can be studied using positive- and negative-sequence components. A common ratio uses negative-sequence magnitude divided by positive-sequence magnitude. Specify the measurement definition rather than treating every imbalance metric as equivalent.", "teacher": "Qwen/Qwen2.5-0.5B-Instruct", "teacher_revision": "7ae557604adf67be50417f59c2c2f167def9a775"} | |
| {"id": "topic-38-style-3", "topic_id": 38, "split": "validation", "prompt": "Give a brief technical explanation of three-phase voltage imbalance. State an important limitation.", "source": "original Sofia prompt bank v1", "response": "Three-phase voltage imbalance can be studied using positive- and negative-sequence components. A common ratio uses negative-sequence magnitude divided by positive-sequence magnitude. Specify the measurement definition rather than treating every imbalance metric as equivalent.", "teacher": "Qwen/Qwen2.5-0.5B-Instruct", "teacher_revision": "7ae557604adf67be50417f59c2c2f167def9a775"} | |
| {"id": "topic-39-style-0", "topic_id": 39, "split": "train", "prompt": "Explain thermal sensor drift and trend analysis in three concise sentences for a junior engineer.", "source": "original Sofia prompt bank v1", "response": "Thermal sensor drift changes the measurement response over time. Compare against reference measurements and operating conditions before interpreting a trend as equipment degradation. Temperature rate estimates are sensitive to sample interval and noise.", "teacher": "Qwen/Qwen2.5-0.5B-Instruct", "teacher_revision": "7ae557604adf67be50417f59c2c2f167def9a775"} | |
| {"id": "topic-39-style-1", "topic_id": 39, "split": "train", "prompt": "What is thermal sensor drift and trend analysis? Give a short practical example.", "source": "original Sofia prompt bank v1", "response": "Thermal sensor drift changes the measurement response over time. Compare against reference measurements and operating conditions before interpreting a trend as equipment degradation. Temperature rate estimates are sensitive to sample interval and noise.", "teacher": "Qwen/Qwen2.5-0.5B-Instruct", "teacher_revision": "7ae557604adf67be50417f59c2c2f167def9a775"} | |
| {"id": "topic-39-style-2", "topic_id": 39, "split": "train", "prompt": "Describe one common mistake involving thermal sensor drift and trend analysis and how to avoid it.", "source": "original Sofia prompt bank v1", "response": "Thermal sensor drift changes the measurement response over time. Compare against reference measurements and operating conditions before interpreting a trend as equipment degradation. Temperature rate estimates are sensitive to sample interval and noise.", "teacher": "Qwen/Qwen2.5-0.5B-Instruct", "teacher_revision": "7ae557604adf67be50417f59c2c2f167def9a775"} | |
| {"id": "topic-39-style-3", "topic_id": 39, "split": "train", "prompt": "Give a brief technical explanation of thermal sensor drift and trend analysis. State an important limitation.", "source": "original Sofia prompt bank v1", "response": "Thermal sensor drift changes the measurement response over time. Compare against reference measurements and operating conditions before interpreting a trend as equipment degradation. Temperature rate estimates are sensitive to sample interval and noise.", "teacher": "Qwen/Qwen2.5-0.5B-Instruct", "teacher_revision": "7ae557604adf67be50417f59c2c2f167def9a775"} | |
| {"id": "topic-40-style-0", "topic_id": 40, "split": "train", "prompt": "Explain the exact classical formula for a product-rotation quantum kernel in three concise sentences for a junior engineer.", "source": "original Sofia prompt bank v1", "response": "For product Ry angle encoding, the squared state-overlap kernel is product_j cos(s*(x_j-z_j)/2)**2. This expression is evaluated exactly on a classical computer in linear work per feature dimension. A useful classification score does not establish quantum computational advantage.", "teacher": "Qwen/Qwen2.5-0.5B-Instruct", "teacher_revision": "7ae557604adf67be50417f59c2c2f167def9a775"} | |
| {"id": "topic-40-style-1", "topic_id": 40, "split": "train", "prompt": "What is the exact classical formula for a product-rotation quantum kernel? Give a short practical example.", "source": "original Sofia prompt bank v1", "response": "For product Ry angle encoding, the squared state-overlap kernel is product_j cos(s*(x_j-z_j)/2)**2. This expression is evaluated exactly on a classical computer in linear work per feature dimension. A useful classification score does not establish quantum computational advantage.", "teacher": "Qwen/Qwen2.5-0.5B-Instruct", "teacher_revision": "7ae557604adf67be50417f59c2c2f167def9a775"} | |
| {"id": "topic-40-style-2", "topic_id": 40, "split": "train", "prompt": "Describe one common mistake involving the exact classical formula for a product-rotation quantum kernel and how to avoid it.", "source": "original Sofia prompt bank v1", "response": "For product Ry angle encoding, the squared state-overlap kernel is product_j cos(s*(x_j-z_j)/2)**2. This expression is evaluated exactly on a classical computer in linear work per feature dimension. A useful classification score does not establish quantum computational advantage.", "teacher": "Qwen/Qwen2.5-0.5B-Instruct", "teacher_revision": "7ae557604adf67be50417f59c2c2f167def9a775"} | |
| {"id": "topic-40-style-3", "topic_id": 40, "split": "train", "prompt": "Give a brief technical explanation of the exact classical formula for a product-rotation quantum kernel. State an important limitation.", "source": "original Sofia prompt bank v1", "response": "For product Ry angle encoding, the squared state-overlap kernel is product_j cos(s*(x_j-z_j)/2)**2. This expression is evaluated exactly on a classical computer in linear work per feature dimension. A useful classification score does not establish quantum computational advantage.", "teacher": "Qwen/Qwen2.5-0.5B-Instruct", "teacher_revision": "7ae557604adf67be50417f59c2c2f167def9a775"} | |
| {"id": "topic-41-style-0", "topic_id": 41, "split": "train", "prompt": "Explain shared terminal-unitary invariance of state overlaps in three concise sentences for a junior engineer.", "source": "original Sofia prompt bank v1", "response": "Applying the same input-independent terminal unitary to both states preserves their inner product. It can change entanglement while leaving the overlap kernel unchanged. An input-dependent layer between operations is outside that cancellation argument.", "teacher": "Qwen/Qwen2.5-0.5B-Instruct", "teacher_revision": "7ae557604adf67be50417f59c2c2f167def9a775"} | |
| {"id": "topic-41-style-1", "topic_id": 41, "split": "train", "prompt": "What is shared terminal-unitary invariance of state overlaps? Give a short practical example.", "source": "original Sofia prompt bank v1", "response": "Applying the same input-independent terminal unitary to both states preserves their inner product. It can change entanglement while leaving the overlap kernel unchanged. An input-dependent layer between operations is outside that cancellation argument.", "teacher": "Qwen/Qwen2.5-0.5B-Instruct", "teacher_revision": "7ae557604adf67be50417f59c2c2f167def9a775"} | |
| {"id": "topic-41-style-2", "topic_id": 41, "split": "train", "prompt": "Describe one common mistake involving shared terminal-unitary invariance of state overlaps and how to avoid it.", "source": "original Sofia prompt bank v1", "response": "Applying the same input-independent terminal unitary to both states preserves their inner product. It can change entanglement while leaving the overlap kernel unchanged. An input-dependent layer between operations is outside that cancellation argument.", "teacher": "Qwen/Qwen2.5-0.5B-Instruct", "teacher_revision": "7ae557604adf67be50417f59c2c2f167def9a775"} | |
| {"id": "topic-41-style-3", "topic_id": 41, "split": "train", "prompt": "Give a brief technical explanation of shared terminal-unitary invariance of state overlaps. State an important limitation.", "source": "original Sofia prompt bank v1", "response": "Applying the same input-independent terminal unitary to both states preserves their inner product. It can change entanglement while leaving the overlap kernel unchanged. An input-dependent layer between operations is outside that cancellation argument.", "teacher": "Qwen/Qwen2.5-0.5B-Instruct", "teacher_revision": "7ae557604adf67be50417f59c2c2f167def9a775"} | |
| {"id": "topic-42-style-0", "topic_id": 42, "split": "train", "prompt": "Explain global depolarization and affine Hilbert-Schmidt kernels in three concise sentences for a junior engineer.", "source": "original Sofia prompt bank v1", "response": "Under common global depolarization of normalized states, the Hilbert-Schmidt kernel becomes a*k+b, with a=(1-p)**2 and b=(1-a)/D. The noisy diagonal is generally below one. This identity does not model arbitrary gate-local noise or Uhlmann fidelity.", "teacher": "Qwen/Qwen2.5-0.5B-Instruct", "teacher_revision": "7ae557604adf67be50417f59c2c2f167def9a775"} | |
| {"id": "topic-42-style-1", "topic_id": 42, "split": "train", "prompt": "What is global depolarization and affine Hilbert-Schmidt kernels? Give a short practical example.", "source": "original Sofia prompt bank v1", "response": "Under common global depolarization of normalized states, the Hilbert-Schmidt kernel becomes a*k+b, with a=(1-p)**2 and b=(1-a)/D. The noisy diagonal is generally below one. This identity does not model arbitrary gate-local noise or Uhlmann fidelity.", "teacher": "Qwen/Qwen2.5-0.5B-Instruct", "teacher_revision": "7ae557604adf67be50417f59c2c2f167def9a775"} | |
| {"id": "topic-42-style-2", "topic_id": 42, "split": "train", "prompt": "Describe one common mistake involving global depolarization and affine Hilbert-Schmidt kernels and how to avoid it.", "source": "original Sofia prompt bank v1", "response": "Under common global depolarization of normalized states, the Hilbert-Schmidt kernel becomes a*k+b, with a=(1-p)**2 and b=(1-a)/D. The noisy diagonal is generally below one. This identity does not model arbitrary gate-local noise or Uhlmann fidelity.", "teacher": "Qwen/Qwen2.5-0.5B-Instruct", "teacher_revision": "7ae557604adf67be50417f59c2c2f167def9a775"} | |
| {"id": "topic-42-style-3", "topic_id": 42, "split": "train", "prompt": "Give a brief technical explanation of global depolarization and affine Hilbert-Schmidt kernels. State an important limitation.", "source": "original Sofia prompt bank v1", "response": "Under common global depolarization of normalized states, the Hilbert-Schmidt kernel becomes a*k+b, with a=(1-p)**2 and b=(1-a)/D. The noisy diagonal is generally below one. This identity does not model arbitrary gate-local noise or Uhlmann fidelity.", "teacher": "Qwen/Qwen2.5-0.5B-Instruct", "teacher_revision": "7ae557604adf67be50417f59c2c2f167def9a775"} | |
| {"id": "topic-43-style-0", "topic_id": 43, "split": "train", "prompt": "Explain matched ridge regularization under affine kernel rescaling in three concise sentences for a junior engineer.", "source": "original Sofia prompt bank v1", "response": "Training-only centering removes the constant offset of an affine kernel transform. For a positive scale a, centred ridge predictions are preserved by scaling the ridge penalty to a*lambda. The completely depolarized endpoint a=0 does not support that inverse equivalence.", "teacher": "Qwen/Qwen2.5-0.5B-Instruct", "teacher_revision": "7ae557604adf67be50417f59c2c2f167def9a775"} | |
| {"id": "topic-43-style-1", "topic_id": 43, "split": "train", "prompt": "What is matched ridge regularization under affine kernel rescaling? Give a short practical example.", "source": "original Sofia prompt bank v1", "response": "Training-only centering removes the constant offset of an affine kernel transform. For a positive scale a, centred ridge predictions are preserved by scaling the ridge penalty to a*lambda. The completely depolarized endpoint a=0 does not support that inverse equivalence.", "teacher": "Qwen/Qwen2.5-0.5B-Instruct", "teacher_revision": "7ae557604adf67be50417f59c2c2f167def9a775"} | |
| {"id": "topic-43-style-2", "topic_id": 43, "split": "train", "prompt": "Describe one common mistake involving matched ridge regularization under affine kernel rescaling and how to avoid it.", "source": "original Sofia prompt bank v1", "response": "Training-only centering removes the constant offset of an affine kernel transform. For a positive scale a, centred ridge predictions are preserved by scaling the ridge penalty to a*lambda. The completely depolarized endpoint a=0 does not support that inverse equivalence.", "teacher": "Qwen/Qwen2.5-0.5B-Instruct", "teacher_revision": "7ae557604adf67be50417f59c2c2f167def9a775"} | |
| {"id": "topic-43-style-3", "topic_id": 43, "split": "train", "prompt": "Give a brief technical explanation of matched ridge regularization under affine kernel rescaling. State an important limitation.", "source": "original Sofia prompt bank v1", "response": "Training-only centering removes the constant offset of an affine kernel transform. For a positive scale a, centred ridge predictions are preserved by scaling the ridge penalty to a*lambda. The completely depolarized endpoint a=0 does not support that inverse equivalence.", "teacher": "Qwen/Qwen2.5-0.5B-Instruct", "teacher_revision": "7ae557604adf67be50417f59c2c2f167def9a775"} | |
| {"id": "topic-44-style-0", "topic_id": 44, "split": "train", "prompt": "Explain finite-shot SWAP measurement uncertainty in three concise sentences for a junior engineer.", "source": "original Sofia prompt bank v1", "response": "An ideal SWAP estimator uses binomial counts with probability (1+k)/2 and estimate 2*C/shots-1. Its variance is (1-k*k)/shots, so negative estimates can occur. Finite measurement noise does not inherit exact matched-regularization invariance.", "teacher": "Qwen/Qwen2.5-0.5B-Instruct", "teacher_revision": "7ae557604adf67be50417f59c2c2f167def9a775"} | |
| {"id": "topic-44-style-1", "topic_id": 44, "split": "train", "prompt": "What is finite-shot SWAP measurement uncertainty? Give a short practical example.", "source": "original Sofia prompt bank v1", "response": "An ideal SWAP estimator uses binomial counts with probability (1+k)/2 and estimate 2*C/shots-1. Its variance is (1-k*k)/shots, so negative estimates can occur. Finite measurement noise does not inherit exact matched-regularization invariance.", "teacher": "Qwen/Qwen2.5-0.5B-Instruct", "teacher_revision": "7ae557604adf67be50417f59c2c2f167def9a775"} | |
| {"id": "topic-44-style-2", "topic_id": 44, "split": "train", "prompt": "Describe one common mistake involving finite-shot SWAP measurement uncertainty and how to avoid it.", "source": "original Sofia prompt bank v1", "response": "An ideal SWAP estimator uses binomial counts with probability (1+k)/2 and estimate 2*C/shots-1. Its variance is (1-k*k)/shots, so negative estimates can occur. Finite measurement noise does not inherit exact matched-regularization invariance.", "teacher": "Qwen/Qwen2.5-0.5B-Instruct", "teacher_revision": "7ae557604adf67be50417f59c2c2f167def9a775"} | |
| {"id": "topic-44-style-3", "topic_id": 44, "split": "train", "prompt": "Give a brief technical explanation of finite-shot SWAP measurement uncertainty. State an important limitation.", "source": "original Sofia prompt bank v1", "response": "An ideal SWAP estimator uses binomial counts with probability (1+k)/2 and estimate 2*C/shots-1. Its variance is (1-k*k)/shots, so negative estimates can occur. Finite measurement noise does not inherit exact matched-regularization invariance.", "teacher": "Qwen/Qwen2.5-0.5B-Instruct", "teacher_revision": "7ae557604adf67be50417f59c2c2f167def9a775"} | |
| {"id": "topic-45-style-0", "topic_id": 45, "split": "train", "prompt": "Explain the difference between useful kernel geometry and quantum computational advantage in three concise sentences for a junior engineer.", "source": "original Sofia prompt bank v1", "response": "A kernel can improve prediction through its feature geometry while remaining efficiently classically evaluable. Compare it against its exact classical implementation and fair competing baselines. Quantum computational advantage needs separate resource and hardware evidence.", "teacher": "Qwen/Qwen2.5-0.5B-Instruct", "teacher_revision": "7ae557604adf67be50417f59c2c2f167def9a775"} | |
| {"id": "topic-45-style-1", "topic_id": 45, "split": "train", "prompt": "What is the difference between useful kernel geometry and quantum computational advantage? Give a short practical example.", "source": "original Sofia prompt bank v1", "response": "A kernel can improve prediction through its feature geometry while remaining efficiently classically evaluable. Compare it against its exact classical implementation and fair competing baselines. Quantum computational advantage needs separate resource and hardware evidence.", "teacher": "Qwen/Qwen2.5-0.5B-Instruct", "teacher_revision": "7ae557604adf67be50417f59c2c2f167def9a775"} | |
| {"id": "topic-45-style-2", "topic_id": 45, "split": "train", "prompt": "Describe one common mistake involving the difference between useful kernel geometry and quantum computational advantage and how to avoid it.", "source": "original Sofia prompt bank v1", "response": "A kernel can improve prediction through its feature geometry while remaining efficiently classically evaluable. Compare it against its exact classical implementation and fair competing baselines. Quantum computational advantage needs separate resource and hardware evidence.", "teacher": "Qwen/Qwen2.5-0.5B-Instruct", "teacher_revision": "7ae557604adf67be50417f59c2c2f167def9a775"} | |
| {"id": "topic-45-style-3", "topic_id": 45, "split": "train", "prompt": "Give a brief technical explanation of the difference between useful kernel geometry and quantum computational advantage. State an important limitation.", "source": "original Sofia prompt bank v1", "response": "A kernel can improve prediction through its feature geometry while remaining efficiently classically evaluable. Compare it against its exact classical implementation and fair competing baselines. Quantum computational advantage needs separate resource and hardware evidence.", "teacher": "Qwen/Qwen2.5-0.5B-Instruct", "teacher_revision": "7ae557604adf67be50417f59c2c2f167def9a775"} | |
| {"id": "topic-46-style-0", "topic_id": 46, "split": "train", "prompt": "Explain training-only centering in kernel ridge regression in three concise sentences for a junior engineer.", "source": "original Sofia prompt bank v1", "response": "Center the training Gram matrix using only training row and grand means. Center each new similarity row using those training statistics and that row's own mean. Estimating an eigenspace from held-out test data would leak evaluation information.", "teacher": "Qwen/Qwen2.5-0.5B-Instruct", "teacher_revision": "7ae557604adf67be50417f59c2c2f167def9a775"} | |
| {"id": "topic-46-style-1", "topic_id": 46, "split": "train", "prompt": "What is training-only centering in kernel ridge regression? Give a short practical example.", "source": "original Sofia prompt bank v1", "response": "Center the training Gram matrix using only training row and grand means. Center each new similarity row using those training statistics and that row's own mean. Estimating an eigenspace from held-out test data would leak evaluation information.", "teacher": "Qwen/Qwen2.5-0.5B-Instruct", "teacher_revision": "7ae557604adf67be50417f59c2c2f167def9a775"} | |
| {"id": "topic-46-style-2", "topic_id": 46, "split": "train", "prompt": "Describe one common mistake involving training-only centering in kernel ridge regression and how to avoid it.", "source": "original Sofia prompt bank v1", "response": "Center the training Gram matrix using only training row and grand means. Center each new similarity row using those training statistics and that row's own mean. Estimating an eigenspace from held-out test data would leak evaluation information.", "teacher": "Qwen/Qwen2.5-0.5B-Instruct", "teacher_revision": "7ae557604adf67be50417f59c2c2f167def9a775"} | |
| {"id": "topic-46-style-3", "topic_id": 46, "split": "train", "prompt": "Give a brief technical explanation of training-only centering in kernel ridge regression. State an important limitation.", "source": "original Sofia prompt bank v1", "response": "Center the training Gram matrix using only training row and grand means. Center each new similarity row using those training statistics and that row's own mean. Estimating an eigenspace from held-out test data would leak evaluation information.", "teacher": "Qwen/Qwen2.5-0.5B-Instruct", "teacher_revision": "7ae557604adf67be50417f59c2c2f167def9a775"} | |
| {"id": "topic-47-style-0", "topic_id": 47, "split": "train", "prompt": "Explain why numerical identity checks do not prove hardware quantum advantage in three concise sentences for a junior engineer.", "source": "original Sofia prompt bank v1", "response": "Numerical identity checks compare independently implemented calculations at chosen inputs. They support implementation correctness but do not replace universal proofs or physical device validation. Report tolerances, tested dimensions, seeds and retained outputs.", "teacher": "Qwen/Qwen2.5-0.5B-Instruct", "teacher_revision": "7ae557604adf67be50417f59c2c2f167def9a775"} | |
| {"id": "topic-47-style-1", "topic_id": 47, "split": "train", "prompt": "What is why numerical identity checks do not prove hardware quantum advantage? Give a short practical example.", "source": "original Sofia prompt bank v1", "response": "Numerical identity checks compare independently implemented calculations at chosen inputs. They support implementation correctness but do not replace universal proofs or physical device validation. Report tolerances, tested dimensions, seeds and retained outputs.", "teacher": "Qwen/Qwen2.5-0.5B-Instruct", "teacher_revision": "7ae557604adf67be50417f59c2c2f167def9a775"} | |
| {"id": "topic-47-style-2", "topic_id": 47, "split": "train", "prompt": "Describe one common mistake involving why numerical identity checks do not prove hardware quantum advantage and how to avoid it.", "source": "original Sofia prompt bank v1", "response": "Numerical identity checks compare independently implemented calculations at chosen inputs. They support implementation correctness but do not replace universal proofs or physical device validation. Report tolerances, tested dimensions, seeds and retained outputs.", "teacher": "Qwen/Qwen2.5-0.5B-Instruct", "teacher_revision": "7ae557604adf67be50417f59c2c2f167def9a775"} | |
| {"id": "topic-47-style-3", "topic_id": 47, "split": "train", "prompt": "Give a brief technical explanation of why numerical identity checks do not prove hardware quantum advantage. State an important limitation.", "source": "original Sofia prompt bank v1", "response": "Numerical identity checks compare independently implemented calculations at chosen inputs. They support implementation correctness but do not replace universal proofs or physical device validation. Report tolerances, tested dimensions, seeds and retained outputs.", "teacher": "Qwen/Qwen2.5-0.5B-Instruct", "teacher_revision": "7ae557604adf67be50417f59c2c2f167def9a775"} | |