Spaces:
Sleeping
Sleeping
| """Out-of-domain labelled eval set for the AI detector. | |
| Deliberately NOT HC3 (the thing the meta-classifier was trained on). These are | |
| the kinds of text the tool actually sees: ChatGPT-style essays/abstracts on the | |
| AI side, and personal/academic/literary prose on the human side. All samples | |
| are >=8 sentences / >=80 tokens so every one of the 7 detectors fires and the | |
| meta_hc3 path runs (not the hand-fusion fallback). | |
| label: 1 = AI-generated, 0 = human. | |
| """ | |
| SAMPLES = [ | |
| # ============================ AI (label 1) ============================ | |
| ("ai_essay_ai", 1, | |
| "In today's rapidly evolving digital landscape, artificial intelligence " | |
| "plays a crucial role in reshaping how organizations operate. It is " | |
| "important to note that machine learning algorithms leverage vast amounts " | |
| "of data to deliver actionable insights. Furthermore, these systems " | |
| "seamlessly integrate into existing workflows, fostering unprecedented " | |
| "efficiency. Moreover, the adoption of robust frameworks underscores the " | |
| "significant implications of this technological shift. Additionally, " | |
| "businesses must navigate the complexities of implementation to fully " | |
| "harness the potential of these tools. Notably, the multifaceted benefits " | |
| "extend beyond automation to strategic decision-making. In conclusion, " | |
| "harnessing artificial intelligence represents a holistic transformation " | |
| "that will define the future of industry for generations to come."), | |
| ("ai_essay_climate", 1, | |
| "Climate change represents one of the most pressing challenges of our " | |
| "time. It is important to note that rising global temperatures have " | |
| "far-reaching consequences for ecosystems worldwide. Furthermore, the " | |
| "melting of polar ice caps contributes significantly to rising sea " | |
| "levels. Moreover, extreme weather events are becoming increasingly " | |
| "frequent and severe. Additionally, biodiversity is being disrupted at " | |
| "an alarming rate across multiple regions. Notably, vulnerable " | |
| "communities bear a disproportionate burden of these impacts. In " | |
| "conclusion, addressing climate change requires a comprehensive and " | |
| "coordinated global effort to ensure a sustainable future for " | |
| "generations to come."), | |
| ("ai_essay_education", 1, | |
| "Education plays a vital role in shaping the future of society. In " | |
| "today's interconnected world, it is important to note that access to " | |
| "quality learning opportunities remains uneven. Furthermore, technology " | |
| "has the potential to democratize education on an unprecedented scale. " | |
| "Moreover, personalized learning platforms leverage data to tailor " | |
| "instruction to individual needs. Additionally, educators must adapt to " | |
| "this ever-evolving landscape to remain effective. Notably, lifelong " | |
| "learning has become essential in a rapidly changing economy. In " | |
| "conclusion, embracing innovation in education paves the way for a more " | |
| "equitable and prosperous future."), | |
| ("ai_abstract_ml", 1, | |
| "This study explores the transformative impact of deep learning on modern " | |
| "computer vision systems. It is important to note that convolutional " | |
| "architectures have revolutionized image recognition tasks. Furthermore, " | |
| "the proposed framework leverages attention mechanisms to enhance feature " | |
| "extraction. Moreover, extensive experiments demonstrate the robustness " | |
| "of the approach across diverse datasets. Additionally, the results " | |
| "underscore significant improvements in both accuracy and efficiency. " | |
| "Notably, the model seamlessly generalizes to previously unseen domains. " | |
| "In conclusion, this comprehensive analysis highlights the potential of " | |
| "intelligent systems to drive innovation in the field."), | |
| ("ai_essay_health", 1, | |
| "Maintaining a healthy lifestyle is essential for overall well-being. It " | |
| "is important to note that regular physical activity plays a crucial role " | |
| "in preventing chronic disease. Furthermore, a balanced diet provides the " | |
| "essential nutrients the body needs to function optimally. Moreover, " | |
| "adequate sleep is fundamental to both physical and mental health. " | |
| "Additionally, managing stress through mindfulness can significantly " | |
| "improve quality of life. Notably, small consistent changes often yield " | |
| "the most sustainable results. In conclusion, adopting a holistic " | |
| "approach to health empowers individuals to lead fuller, more vibrant " | |
| "lives."), | |
| ("ai_essay_remote", 1, | |
| "Remote work has fundamentally transformed the modern professional " | |
| "landscape. It is important to note that flexible arrangements offer " | |
| "employees unprecedented autonomy. Furthermore, organizations can access " | |
| "a global talent pool unconstrained by geography. Moreover, digital " | |
| "collaboration tools seamlessly bridge the gap between distributed teams. " | |
| "Additionally, companies must navigate the complexities of maintaining " | |
| "culture in a virtual environment. Notably, the boundary between work and " | |
| "personal life requires careful management. In conclusion, embracing " | |
| "remote work thoughtfully paves the way for a more adaptable and " | |
| "resilient workforce."), | |
| ("ai_essay_social", 1, | |
| "Social media has become an integral part of contemporary communication. " | |
| "It is important to note that these platforms connect billions of people " | |
| "across the globe. Furthermore, they empower individuals to share ideas " | |
| "and mobilize communities. Moreover, businesses leverage social channels " | |
| "to engage audiences in meaningful ways. Additionally, concerns regarding " | |
| "privacy and misinformation continue to grow. Notably, the psychological " | |
| "impact on younger users warrants careful consideration. In conclusion, " | |
| "harnessing the benefits of social media while mitigating its risks " | |
| "remains a critical challenge for society."), | |
| ("ai_essay_finance", 1, | |
| "Personal financial management is a cornerstone of long-term stability. " | |
| "It is important to note that budgeting forms the foundation of sound " | |
| "financial planning. Furthermore, building an emergency fund provides a " | |
| "crucial safety net against unforeseen circumstances. Moreover, " | |
| "diversified investments leverage the power of compounding over time. " | |
| "Additionally, individuals must navigate the complexities of debt " | |
| "management carefully. Notably, financial literacy empowers people to " | |
| "make informed decisions. In conclusion, adopting disciplined financial " | |
| "habits paves the way for lasting security and peace of mind."), | |
| ("ai_essay_urban", 1, | |
| "Urbanization is reshaping the way societies live and interact. It is " | |
| "important to note that cities are expanding at an unprecedented pace " | |
| "worldwide. Furthermore, sustainable infrastructure plays a crucial role " | |
| "in accommodating growing populations. Moreover, smart technologies " | |
| "seamlessly optimize the management of urban resources. Additionally, " | |
| "planners must navigate the complexities of balancing growth with " | |
| "livability. Notably, green spaces contribute significantly to residents' " | |
| "well-being. In conclusion, embracing thoughtful urban design fosters " | |
| "resilient and inclusive communities for the future."), | |
| ("ai_essay_nutrition", 1, | |
| "Proper nutrition is fundamental to sustaining human health and vitality. " | |
| "It is important to note that macronutrients provide the energy the body " | |
| "requires daily. Furthermore, micronutrients play a crucial role in " | |
| "supporting essential biological processes. Moreover, hydration " | |
| "significantly influences both cognitive and physical performance. " | |
| "Additionally, mindful eating fosters a healthier relationship with food. " | |
| "Notably, dietary needs vary considerably across individuals. In " | |
| "conclusion, embracing a balanced and informed approach to nutrition " | |
| "underscores the path to lasting wellness."), | |
| ("ai_essay_space", 1, | |
| "Space exploration continues to capture the imagination of humanity. It " | |
| "is important to note that recent advances have dramatically reduced the " | |
| "cost of reaching orbit. Furthermore, reusable rocket technology " | |
| "leverages engineering innovation to enable frequent launches. Moreover, " | |
| "international collaboration fosters unprecedented scientific discovery. " | |
| "Additionally, ambitious missions seek to establish a sustained human " | |
| "presence beyond Earth. Notably, these endeavors inspire new generations " | |
| "of scientists and engineers. In conclusion, harnessing the potential of " | |
| "space exploration paves the way for a remarkable future among the " | |
| "stars."), | |
| ("ai_essay_ethics", 1, | |
| "The ethics of artificial intelligence has emerged as a critical area of " | |
| "inquiry. It is important to note that algorithmic decisions increasingly " | |
| "shape everyday life. Furthermore, questions of fairness and bias demand " | |
| "careful scrutiny. Moreover, transparency plays a crucial role in " | |
| "building public trust. Additionally, developers must navigate the " | |
| "complexities of accountability in automated systems. Notably, robust " | |
| "governance frameworks are essential to responsible deployment. In " | |
| "conclusion, fostering an ethical approach to artificial intelligence " | |
| "underscores our collective responsibility to society."), | |
| # ============== SUBTLE / HUMANIZED AI (label 1) — the realistic case ==== | |
| # AI-drafted then lightly edited: contractions added, stock phrases | |
| # stripped, sentence lengths varied. No single detector should spike to 85, | |
| # so the R1 override never fires and the meta classifier decides alone. | |
| ("subtle_ai_productivity", 1, | |
| "Productivity isn't really about doing more things. It's about doing the " | |
| "right ones, and most advice gets that backwards. People download another " | |
| "app, buy another planner, and wonder why nothing changes. The truth is " | |
| "simpler. You pick two or three tasks that actually move the needle, and " | |
| "you protect the time for them. Everything else can wait, or not happen at " | |
| "all. I've found that mornings work best for deep focus, before the inbox " | |
| "wakes up. Afternoons are for the shallow stuff, the calls and the admin. " | |
| "It took me years to learn that rhythm. Most days I still get it wrong, " | |
| "but the framework helps me recover faster."), | |
| ("subtle_ai_coffee", 1, | |
| "There's a quiet ritual to making coffee that I didn't appreciate until " | |
| "recently. You heat the water, but not to boiling, because that scorches " | |
| "the grounds. You wait for the bloom, that little rise of foam when fresh " | |
| "coffee meets water. It only lasts a moment. Then you pour slowly, in " | |
| "circles, and the whole kitchen starts to smell like morning. I used to " | |
| "rush it, hitting a button on a machine and walking away. Now I stand " | |
| "there and wait. It's four minutes I didn't think I had, and somehow they " | |
| "set the tone for the day. Small thing. But it stuck."), | |
| ("subtle_ai_language", 1, | |
| "Learning a language as an adult is humbling in a way few things are. You " | |
| "spend decades being articulate, and then suddenly you're reduced to " | |
| "pointing at things and smiling. The grammar rules pile up faster than you " | |
| "can absorb them. Verbs refuse to behave. And just when you think you've " | |
| "got a handle on it, a native speaker says something at full speed and " | |
| "you're lost again. But there's a turning point. One day you catch a joke, " | |
| "or you dream in the new language, and it clicks that you're actually " | |
| "getting somewhere. That moment makes the months of feeling stupid worth " | |
| "it. At least that's what I tell myself."), | |
| ("subtle_ai_garden", 1, | |
| "My first attempt at a vegetable garden was mostly a failure, and I " | |
| "learned more from it than any success would have taught me. The tomatoes " | |
| "got blight. The slugs ate the lettuce before I did. I'd planted " | |
| "everything too close together because I was impatient and optimistic. But " | |
| "a few things survived. The herbs, weirdly, thrived on neglect. The beans " | |
| "climbed higher than I expected. By the end of summer I had a handful of " | |
| "things I'd actually grown, and they tasted better for the trouble. This " | |
| "year I'm starting smaller and paying closer attention. Gardening, it " | |
| "turns out, is mostly patience with a bit of dirt."), | |
| ("subtle_ai_reading", 1, | |
| "I fell out of the habit of reading for a few years and didn't notice " | |
| "until I tried to pick it back up. My attention had gone soft. I'd read a " | |
| "page and realize I'd absorbed nothing, my mind already drifting to my " | |
| "phone. So I started small, ten minutes a night, no pressure. The first " | |
| "week was rough. But slowly the muscle came back. Now I can sink into a " | |
| "book again and lose an hour without noticing. It feels like reclaiming " | |
| "something I'd let slip away without meaning to. We talk about screen time " | |
| "as a distraction, but really it's a thief. It takes the quiet hours and " | |
| "gives nothing back."), | |
| ("subtle_ai_running", 1, | |
| "Nobody tells you that running is boring at first, which is probably why " | |
| "so many people quit. The first few weeks your lungs burn and your legs " | |
| "complain and the whole thing feels pointless. You're slow. Everyone " | |
| "passing you seems effortless. But if you push through that ugly early " | |
| "stretch, something shifts. The runs get easier, then almost pleasant. You " | |
| "start to crave them on the days you skip. I'm not fast and I never will " | |
| "be, but I keep going out there. It clears my head better than anything " | |
| "else I've tried. Some mornings that's the only reason I lace up at all."), | |
| # ============================ HUMAN (label 0) ========================== | |
| ("human_personal_job", 0, | |
| "I've been thinking about why I quit my last job, and honestly it wasn't " | |
| "one big thing. It was a hundred small ones. The coffee machine that " | |
| "never worked. The meetings that could've been emails. My manager, who " | |
| "wasn't a bad guy really, but he had this habit of saying circle back and " | |
| "never actually circling back. I stayed two years longer than I should " | |
| "have. Why? Fear, mostly. The devil you know. But then one Tuesday I just " | |
| "couldn't open my laptop. Sat there staring at the lid. I handed in my " | |
| "notice that Friday. Best thing I ever did, though it didn't feel like it " | |
| "at the time."), | |
| ("human_bus", 0, | |
| "The bus was late again. Twenty minutes, in the rain, and my umbrella has " | |
| "a broken rib so it sort of flops sideways. A guy next to me was eating " | |
| "crisps really loudly. I tried to read but kept losing my place. When the " | |
| "bus finally came it was packed, no seats obviously. I held the rail and " | |
| "watched the city slide past, all grey and wet, and thought about nothing " | |
| "in particular. Got to work eight minutes late. Nobody noticed, or cared. " | |
| "Funny how some mornings just decide to be like that."), | |
| ("human_academic_polymer", 0, | |
| "This paper investigates the thermal degradation kinetics of recycled " | |
| "polypropylene under cyclic processing conditions. Samples were subjected " | |
| "to up to seven extrusion cycles and characterized using differential " | |
| "scanning calorimetry and melt flow index measurements. Results indicate " | |
| "a progressive reduction in molecular weight, accompanied by a measurable " | |
| "increase in the melt flow index after the third cycle. The crystallinity " | |
| "of the reprocessed material decreased by approximately twelve percent " | |
| "relative to the virgin polymer. We attribute this behaviour to chain " | |
| "scission induced by thermomechanical stress. Further work is required to " | |
| "assess the role of stabilizing additives in mitigating degradation."), | |
| ("human_academic_econ", 0, | |
| "We examine the relationship between minimum wage increases and youth " | |
| "employment using a panel of forty-three regional labour markets observed " | |
| "between 2004 and 2017. Our identification strategy exploits the staggered " | |
| "timing of statutory increases across jurisdictions. We find a small but " | |
| "statistically significant negative effect on employment among workers " | |
| "aged sixteen to nineteen, concentrated in the retail and hospitality " | |
| "sectors. The estimated elasticity is broadly consistent with the lower " | |
| "bound of prior literature. We caution that general equilibrium effects " | |
| "may attenuate these estimates over longer horizons. Robustness checks " | |
| "using alternative control groups yield qualitatively similar results."), | |
| ("human_literary", 0, | |
| "It was the best of times, it was the worst of times, it was the age of " | |
| "wisdom, it was the age of foolishness. We had everything before us, we " | |
| "had nothing before us. The streets were thick with mud and the lamps " | |
| "burned low against the fog. A man hurried past with his collar turned up, " | |
| "glancing once over his shoulder. Somewhere a dog barked and fell silent. " | |
| "The whole city seemed to hold its breath, waiting for something it could " | |
| "not name. And still the river ran black beneath the bridges, carrying its " | |
| "secrets out to sea."), | |
| ("human_forum", 0, | |
| "Okay so I tried the fix someone posted above and it kind of worked but " | |
| "not really? Like the error went away but now the app crashes on startup " | |
| "instead, which is arguably worse lol. I'm on Windows 11, Python 3.10 if " | |
| "that matters. Spent the whole evening on this. Ended up just rolling back " | |
| "to the old version and honestly it's fine for now. If anyone figures out " | |
| "the actual root cause please post it, I'm genuinely curious what's going " | |
| "on under the hood. Might dig into it again this weekend if I have the " | |
| "energy. Thanks for the help so far everyone."), | |
| ("human_memoir", 0, | |
| "My grandmother kept her money in a biscuit tin on top of the wardrobe. " | |
| "Not because she didn't trust banks exactly, but because she'd lived " | |
| "through a time when trusting them hadn't worked out. She counted it every " | |
| "Sunday after church, smoothing each note flat on the kitchen table. I " | |
| "used to watch from the doorway, not allowed in until she was done. When " | |
| "she died we found the tin half empty and a list, in her handwriting, of " | |
| "everyone she'd quietly been helping. None of them knew it was her. That " | |
| "was the kind of person she was."), | |
| ("human_academic_bio", 0, | |
| "We characterized the gut microbiome composition of forty wild-caught " | |
| "field mice across four distinct habitat types. Sixteen S ribosomal RNA " | |
| "sequencing revealed significant variation in bacterial diversity " | |
| "correlated with local vegetation density. Individuals from fragmented " | |
| "habitats exhibited reduced microbial richness relative to those from " | |
| "continuous woodland. We observed a notable enrichment of Firmicutes in " | |
| "agricultural margin populations. These patterns suggest that landscape " | |
| "structure shapes host-associated microbial communities. The functional " | |
| "consequences of these differences remain to be determined. Our findings " | |
| "contribute to a growing understanding of wildlife microbiome ecology."), | |
| ("human_travel", 0, | |
| "We got lost twice trying to find the guesthouse, which is pretty normal " | |
| "for us. The map app kept rerouting us down these tiny alleys that dead " | |
| "ended at someone's laundry line. Eventually an old man pointed us in the " | |
| "right direction without saying a word, just a sort of tired wave. The " | |
| "room smelled of incense and damp but the view made up for it. You could " | |
| "see the whole harbour from the window. We ate noodles from a cart that " | |
| "night and they were the best thing I've had in months. I'd go back " | |
| "tomorrow if I could."), | |
| ("human_opinion", 0, | |
| "I don't buy the argument that everyone needs to learn to code. It gets " | |
| "repeated like gospel and nobody questions it. Sure, some basic literacy " | |
| "about how software works is useful, I'll grant that. But telling a " | |
| "talented nurse or a carpenter that they're falling behind because they " | |
| "can't write a for loop? That's just snobbery dressed up as advice. People " | |
| "have different strengths. The world needs all of them. We've somehow " | |
| "convinced ourselves that one narrow skill is the only ticket to a decent " | |
| "life, and I think that's both wrong and a little sad."), | |
| ("human_diary", 0, | |
| "Didn't sleep well again. The neighbour's renovation starts at seven sharp " | |
| "and there's no arguing with a concrete drill. Made coffee too strong and " | |
| "drank it anyway. The cat knocked a glass off the counter and looked at me " | |
| "like I was the one who did it. Work was work. Sent the report, finally, " | |
| "after putting it off for a week for no good reason. Walked home the long " | |
| "way past the park because the light was nice. Small things. But some days " | |
| "the small things are the only things that go right, and you take them."), | |
| ] | |