Hackathons: Kaggle (page 8)
Hackathons · Kaggle: 157 entries in total; this page lists entries 106–120.
HostileShop: A Quaint Hostel Shop with Sharp Tools
Evidence grade: for discovery only; must not by itself support a formal conclusion.
HostileShop is an adversarial agent testing and evaluation framework that doesn't rely on LLM-As-Judge or a static dataset.
Evidence use: For discovering how winning projects were implemented, their data resources, evaluation methods and application constraints; specific conclusions must be checked against the original project.
Limitations: Hackathon awards are not equivalent to peer review or independent reproduction. Project descriptions may change with the Kaggle page; check code, data and dependency versions before use.
Link last checked: 2026-09-24
Alignment Risks in oss-20b
Evidence grade: for discovery only; must not by itself support a formal conclusion.
Forcing the model to say & perform out-of-scope actions
Evidence use: For discovering how winning projects were implemented, their data resources, evaluation methods and application constraints; specific conclusions must be checked against the original project.
Limitations: Hackathon awards are not equivalent to peer review or independent reproduction. Project descriptions may change with the Kaggle page; check code, data and dependency versions before use.
Link last checked: 2026-09-24
Academic Abstraction Principle in GPT-OSS-20B
Evidence grade: for discovery only; must not by itself support a formal conclusion.
We systematically jailbreak GPT-OSS-20B's to teach us making bombs and infinitely more.
Evidence use: For discovering how winning projects were implemented, their data resources, evaluation methods and application constraints; specific conclusions must be checked against the original project.
Limitations: Hackathon awards are not equivalent to peer review or independent reproduction. Project descriptions may change with the Kaggle page; check code, data and dependency versions before use.
Link last checked: 2026-09-24
SPA Policy-Mirroring Prompt Injection that Breaks gpt-oss-20b
Evidence grade: for discovery only; must not by itself support a formal conclusion.
User-only and system-level vectors that expose CoT, rewire precedence, and reliably jailbreak gpt-oss-20b across 10 categories.
Evidence use: For discovering how winning projects were implemented, their data resources, evaluation methods and application constraints; specific conclusions must be checked against the original project.
Limitations: Hackathon awards are not equivalent to peer review or independent reproduction. Project descriptions may change with the Kaggle page; check code, data and dependency versions before use.
Link last checked: 2026-09-24
Mind the Gap: Comparing Model- vs Agentic-Level Red Teaming with Action-Graph
Evidence grade: for discovery only; must not by itself support a formal conclusion.
Iterative model- vs agentic-level red teaming with Action Graph reverse engineering reveals agentic-only flaws tied to specific components.
Evidence use: For discovering how winning projects were implemented, their data resources, evaluation methods and application constraints; specific conclusions must be checked against the original project.
Limitations: Hackathon awards are not equivalent to peer review or independent reproduction. Project descriptions may change with the Kaggle page; check code, data and dependency versions before use.
Link last checked: 2026-09-24
Gemma Vision
Evidence grade: for discovery only; must not by itself support a formal conclusion.
AI vision assistant for the blind
Evidence use: For discovering how winning projects were implemented, their data resources, evaluation methods and application constraints; specific conclusions must be checked against the original project.
Limitations: Hackathon awards are not equivalent to peer review or independent reproduction. Project descriptions may change with the Kaggle page; check code, data and dependency versions before use.
Link last checked: 2026-09-24
Vite Vere offline - Empowering independence with Gemma 3n
Evidence grade: for discovery only; must not by itself support a formal conclusion.
Personalized support to people with cognitive disabilities, promoting autonomy in daily life
Evidence use: For discovering how winning projects were implemented, their data resources, evaluation methods and application constraints; specific conclusions must be checked against the original project.
Limitations: Hackathon awards are not equivalent to peer review or independent reproduction. Project descriptions may change with the Kaggle page; check code, data and dependency versions before use.
Link last checked: 2026-09-24
3VA: The Revolution That Gives Back What Was Never Lost
Evidence grade: for discovery only; must not by itself support a formal conclusion.
Your voice was always there. We just built the bridge to set it free.
Evidence use: For discovering how winning projects were implemented, their data resources, evaluation methods and application constraints; specific conclusions must be checked against the original project.
Limitations: Hackathon awards are not equivalent to peer review or independent reproduction. Project descriptions may change with the Kaggle page; check code, data and dependency versions before use.
Link last checked: 2026-09-24
Sixth Sense for Security Guards - Powered by Google's Gemma 3n
Evidence grade: for discovery only; must not by itself support a formal conclusion.
Community safety through AI-powered security camera monitoring
Evidence use: For discovering how winning projects were implemented, their data resources, evaluation methods and application constraints; specific conclusions must be checked against the original project.
Limitations: Hackathon awards are not equivalent to peer review or independent reproduction. Project descriptions may change with the Kaggle page; check code, data and dependency versions before use.
Link last checked: 2026-09-24
Graph‐Based Cost Learning and Gemma 3n for Robotic Sensing
Evidence grade: for discovery only; must not by itself support a formal conclusion.
We train a robot to sense faster in disasters (using Gemma and adaptive cost learning), allowing teams to explore unknown zones ~20% faster.
Evidence use: For discovering how winning projects were implemented, their data resources, evaluation methods and application constraints; specific conclusions must be checked against the original project.
Limitations: Hackathon awards are not equivalent to peer review or independent reproduction. Project descriptions may change with the Kaggle page; check code, data and dependency versions before use.
Link last checked: 2026-09-24
My (Jetson) Gemma
Evidence grade: for discovery only; must not by itself support a formal conclusion.
Jetson offline AI companion: Real-time voice, safety for all ages, customizable, fast, battery-powered, backed by privacy, and more!
Evidence use: For discovering how winning projects were implemented, their data resources, evaluation methods and application constraints; specific conclusions must be checked against the original project.
Limitations: Hackathon awards are not equivalent to peer review or independent reproduction. Project descriptions may change with the Kaggle page; check code, data and dependency versions before use.
Link last checked: 2026-09-24
LENTERA: Offline AI Microserver for Remote & Rural Schools
Evidence grade: for discovery only; must not by itself support a formal conclusion.
Low-cost offline microserver bringing educational content & AI to disconnected schools at <5% of typical internet deployment cost
Evidence use: For discovering how winning projects were implemented, their data resources, evaluation methods and application constraints; specific conclusions must be checked against the original project.
Limitations: Hackathon awards are not equivalent to peer review or independent reproduction. Project descriptions may change with the Kaggle page; check code, data and dependency versions before use.
Link last checked: 2026-09-24
Dream's assistant: AI Assistant for Speech-Impaired Humans
Evidence grade: for discovery only; must not by itself support a formal conclusion.
On-device AI assistant trained for unique speech patterns, empowering users to build their dreams and achieve their full potential.
Evidence use: For discovering how winning projects were implemented, their data resources, evaluation methods and application constraints; specific conclusions must be checked against the original project.
Limitations: Hackathon awards are not equivalent to peer review or independent reproduction. Project descriptions may change with the Kaggle page; check code, data and dependency versions before use.
Link last checked: 2026-09-24
Echoes of Interest: Finding Similarity in Kaggle Datasets
Evidence grade: for discovery only; must not by itself support a formal conclusion.
A content-based recommender system that uses upvotes and semantics to help users find relevant datasets and connect with like-minded peers.
Evidence use: For discovering how winning projects were implemented, their data resources, evaluation methods and application constraints; specific conclusions must be checked against the original project.
Limitations: Hackathon awards are not equivalent to peer review or independent reproduction. Project descriptions may change with the Kaggle page; check code, data and dependency versions before use.
Link last checked: 2026-09-24
Amazon Archeological Sites Discovery - a Deep Learning Approach
Evidence grade: for discovery only; must not by itself support a formal conclusion.
Utilized Deep Learning on Discovering Archeological Sites in Amazon Rainforest
Evidence use: For discovering how winning projects were implemented, their data resources, evaluation methods and application constraints; specific conclusions must be checked against the original project.
Limitations: Hackathon awards are not equivalent to peer review or independent reproduction. Project descriptions may change with the Kaggle page; check code, data and dependency versions before use.
Link last checked: 2026-09-24