@cartathomas.bsky.social
🧭MAGELLAN is built on on many works that use of LP to drive automatic curriculum learning e.g. by @rockt.ai @egrefen.bsky.social @jeffclune.com @tomssilver.bsky.social @tambetm.bsky.social @jrsrichmond.bsky.social @ryanpsullivan.bsky.social
March 24, 2025 at 3:09 PM
🧭MAGELLAN is built on on many works that use of LP to drive automatic curriculum learning e.g. by @rockt.ai @egrefen.bsky.social @jeffclune.com @tomssilver.bsky.social @tambetm.bsky.social @jrsrichmond.bsky.social @ryanpsullivan.bsky.social
Thanks to @lorisgaven.bsky.social @clementromac.bsky.social for the fun time doing research on this topic and huge thanks also to, @ccolas.bsky.social Sylvain Lamprier, Olivier Sigaud and @pyoudeyer.bsky.social for their supervision!!
March 24, 2025 at 3:09 PM
Thanks to @lorisgaven.bsky.social @clementromac.bsky.social for the fun time doing research on this topic and huge thanks also to, @ccolas.bsky.social Sylvain Lamprier, Olivier Sigaud and @pyoudeyer.bsky.social for their supervision!!
𝐐𝟒: 𝐀𝐝𝐚𝐩𝐭𝐚𝐭𝐢𝐨𝐧 𝐭𝐨 𝐄𝐯𝐨𝐥𝐯𝐢𝐧𝐠 𝐆𝐨𝐚𝐥 𝐒𝐩𝐚𝐜𝐞𝐬
We replaced the 𝐞𝐧𝐭𝐢𝐫𝐞 𝐠𝐨𝐚𝐥 𝐬𝐩𝐚𝐜𝐞 with unseen goals from the same categories. 🧭MAGELLAN generalized LP and retained exceptional performance—matching baselines that rely on human expertise! 🚀✨
We replaced the 𝐞𝐧𝐭𝐢𝐫𝐞 𝐠𝐨𝐚𝐥 𝐬𝐩𝐚𝐜𝐞 with unseen goals from the same categories. 🧭MAGELLAN generalized LP and retained exceptional performance—matching baselines that rely on human expertise! 🚀✨
March 24, 2025 at 3:09 PM
𝐐𝟒: 𝐀𝐝𝐚𝐩𝐭𝐚𝐭𝐢𝐨𝐧 𝐭𝐨 𝐄𝐯𝐨𝐥𝐯𝐢𝐧𝐠 𝐆𝐨𝐚𝐥 𝐒𝐩𝐚𝐜𝐞𝐬
We replaced the 𝐞𝐧𝐭𝐢𝐫𝐞 𝐠𝐨𝐚𝐥 𝐬𝐩𝐚𝐜𝐞 with unseen goals from the same categories. 🧭MAGELLAN generalized LP and retained exceptional performance—matching baselines that rely on human expertise! 🚀✨
We replaced the 𝐞𝐧𝐭𝐢𝐫𝐞 𝐠𝐨𝐚𝐥 𝐬𝐩𝐚𝐜𝐞 with unseen goals from the same categories. 🧭MAGELLAN generalized LP and retained exceptional performance—matching baselines that rely on human expertise! 🚀✨
𝐐𝟑: 𝐆𝐞𝐧𝐞𝐫𝐚𝐥𝐢𝐳𝐚𝐭𝐢𝐨𝐧
At the end of training, 🧭MAGELLAN has 𝐬𝐭𝐫𝐮𝐜𝐭𝐮𝐫𝐞𝐝 𝐭𝐡𝐞 𝐠𝐨𝐚𝐥 𝐞𝐦𝐛𝐞𝐝𝐝𝐢𝐧𝐠 𝐬𝐩𝐚𝐜𝐞, consistently 𝐩𝐫𝐞𝐝𝐢𝐜𝐭𝐢𝐧𝐠 success probability 𝐟𝐨𝐫 𝐮𝐧𝐬𝐞𝐞𝐧 𝐠𝐨𝐚𝐥𝐬, a key step toward scalable open-ended learning!
At the end of training, 🧭MAGELLAN has 𝐬𝐭𝐫𝐮𝐜𝐭𝐮𝐫𝐞𝐝 𝐭𝐡𝐞 𝐠𝐨𝐚𝐥 𝐞𝐦𝐛𝐞𝐝𝐝𝐢𝐧𝐠 𝐬𝐩𝐚𝐜𝐞, consistently 𝐩𝐫𝐞𝐝𝐢𝐜𝐭𝐢𝐧𝐠 success probability 𝐟𝐨𝐫 𝐮𝐧𝐬𝐞𝐞𝐧 𝐠𝐨𝐚𝐥𝐬, a key step toward scalable open-ended learning!
March 24, 2025 at 3:09 PM
𝐐𝟑: 𝐆𝐞𝐧𝐞𝐫𝐚𝐥𝐢𝐳𝐚𝐭𝐢𝐨𝐧
At the end of training, 🧭MAGELLAN has 𝐬𝐭𝐫𝐮𝐜𝐭𝐮𝐫𝐞𝐝 𝐭𝐡𝐞 𝐠𝐨𝐚𝐥 𝐞𝐦𝐛𝐞𝐝𝐝𝐢𝐧𝐠 𝐬𝐩𝐚𝐜𝐞, consistently 𝐩𝐫𝐞𝐝𝐢𝐜𝐭𝐢𝐧𝐠 success probability 𝐟𝐨𝐫 𝐮𝐧𝐬𝐞𝐞𝐧 𝐠𝐨𝐚𝐥𝐬, a key step toward scalable open-ended learning!
At the end of training, 🧭MAGELLAN has 𝐬𝐭𝐫𝐮𝐜𝐭𝐮𝐫𝐞𝐝 𝐭𝐡𝐞 𝐠𝐨𝐚𝐥 𝐞𝐦𝐛𝐞𝐝𝐝𝐢𝐧𝐠 𝐬𝐩𝐚𝐜𝐞, consistently 𝐩𝐫𝐞𝐝𝐢𝐜𝐭𝐢𝐧𝐠 success probability 𝐟𝐨𝐫 𝐮𝐧𝐬𝐞𝐞𝐧 𝐠𝐨𝐚𝐥𝐬, a key step toward scalable open-ended learning!
𝐐𝟐: 𝐂𝐮𝐫𝐫𝐢𝐜𝐮𝐥𝐮𝐦 𝐋𝐞𝐚𝐫𝐧𝐢𝐧𝐠
🧭MAGELLAN autonomously discovers goal families (✊🌿🐮🦁) across 𝟐𝟓𝐤 𝐠𝐨𝐚𝐥𝐬, performing on par with expert knowledge-augmented baselines—but 𝐰𝐢𝐭𝐡𝐨𝐮𝐭 𝐫𝐞𝐪𝐮𝐢𝐫𝐢𝐧𝐠 𝐩𝐫𝐞𝐝𝐞𝐟𝐢𝐧𝐞𝐝 𝐠𝐨𝐚𝐥 𝐜𝐥𝐮𝐬𝐭𝐞𝐫𝐬! 🚀
🧭MAGELLAN autonomously discovers goal families (✊🌿🐮🦁) across 𝟐𝟓𝐤 𝐠𝐨𝐚𝐥𝐬, performing on par with expert knowledge-augmented baselines—but 𝐰𝐢𝐭𝐡𝐨𝐮𝐭 𝐫𝐞𝐪𝐮𝐢𝐫𝐢𝐧𝐠 𝐩𝐫𝐞𝐝𝐞𝐟𝐢𝐧𝐞𝐝 𝐠𝐨𝐚𝐥 𝐜𝐥𝐮𝐬𝐭𝐞𝐫𝐬! 🚀
March 24, 2025 at 3:09 PM
𝐐𝟐: 𝐂𝐮𝐫𝐫𝐢𝐜𝐮𝐥𝐮𝐦 𝐋𝐞𝐚𝐫𝐧𝐢𝐧𝐠
🧭MAGELLAN autonomously discovers goal families (✊🌿🐮🦁) across 𝟐𝟓𝐤 𝐠𝐨𝐚𝐥𝐬, performing on par with expert knowledge-augmented baselines—but 𝐰𝐢𝐭𝐡𝐨𝐮𝐭 𝐫𝐞𝐪𝐮𝐢𝐫𝐢𝐧𝐠 𝐩𝐫𝐞𝐝𝐞𝐟𝐢𝐧𝐞𝐝 𝐠𝐨𝐚𝐥 𝐜𝐥𝐮𝐬𝐭𝐞𝐫𝐬! 🚀
🧭MAGELLAN autonomously discovers goal families (✊🌿🐮🦁) across 𝟐𝟓𝐤 𝐠𝐨𝐚𝐥𝐬, performing on par with expert knowledge-augmented baselines—but 𝐰𝐢𝐭𝐡𝐨𝐮𝐭 𝐫𝐞𝐪𝐮𝐢𝐫𝐢𝐧𝐠 𝐩𝐫𝐞𝐝𝐞𝐟𝐢𝐧𝐞𝐝 𝐠𝐨𝐚𝐥 𝐜𝐥𝐮𝐬𝐭𝐞𝐫𝐬! 🚀
🎯 𝐐𝟏: 𝐂𝐨𝐦𝐩𝐞𝐭𝐞𝐧𝐜𝐞 𝐄𝐬𝐭𝐢𝐦𝐚𝐭𝐢𝐨𝐧
🧭MAGELLAN matches expert baselines in estimating competence over tens of thousands of goals but with 𝐦𝐢𝐧𝐢𝐦𝐚𝐥 𝐜𝐨𝐬𝐭 & 𝐞𝐫𝐫𝐨𝐫! Unlike other methods, it efficiently tracks competence transfer across large goal spaces
🧭MAGELLAN matches expert baselines in estimating competence over tens of thousands of goals but with 𝐦𝐢𝐧𝐢𝐦𝐚𝐥 𝐜𝐨𝐬𝐭 & 𝐞𝐫𝐫𝐨𝐫! Unlike other methods, it efficiently tracks competence transfer across large goal spaces
March 24, 2025 at 3:09 PM
🎯 𝐐𝟏: 𝐂𝐨𝐦𝐩𝐞𝐭𝐞𝐧𝐜𝐞 𝐄𝐬𝐭𝐢𝐦𝐚𝐭𝐢𝐨𝐧
🧭MAGELLAN matches expert baselines in estimating competence over tens of thousands of goals but with 𝐦𝐢𝐧𝐢𝐦𝐚𝐥 𝐜𝐨𝐬𝐭 & 𝐞𝐫𝐫𝐨𝐫! Unlike other methods, it efficiently tracks competence transfer across large goal spaces
🧭MAGELLAN matches expert baselines in estimating competence over tens of thousands of goals but with 𝐦𝐢𝐧𝐢𝐦𝐚𝐥 𝐜𝐨𝐬𝐭 & 𝐞𝐫𝐫𝐨𝐫! Unlike other methods, it efficiently tracks competence transfer across large goal spaces
We studied 4 scientific questions:
Q1 How does 🧭MAGELLAN's competence estimation compare to classical approaches?
Q2 Can it be used to build an efficient curriculum?
Q3 Can it generalize on unseen goals?
Q4 Can it adapt to an evolving goal space?
Let's dive in! 👇
Q1 How does 🧭MAGELLAN's competence estimation compare to classical approaches?
Q2 Can it be used to build an efficient curriculum?
Q3 Can it generalize on unseen goals?
Q4 Can it adapt to an evolving goal space?
Let's dive in! 👇
March 24, 2025 at 3:09 PM
We studied 4 scientific questions:
Q1 How does 🧭MAGELLAN's competence estimation compare to classical approaches?
Q2 Can it be used to build an efficient curriculum?
Q3 Can it generalize on unseen goals?
Q4 Can it adapt to an evolving goal space?
Let's dive in! 👇
Q1 How does 🧭MAGELLAN's competence estimation compare to classical approaches?
Q2 Can it be used to build an efficient curriculum?
Q3 Can it generalize on unseen goals?
Q4 Can it adapt to an evolving goal space?
Let's dive in! 👇
By capturing semantic relationships between goals, 🧭MAGELLAN enables efficient 𝐋𝐏 𝐞𝐬𝐭𝐢𝐦𝐚𝐭𝐢𝐨𝐧 & adaptive goal prioritization—all 𝐰𝐢𝐭𝐡𝐨𝐮𝐭 𝐫𝐞𝐥𝐲𝐢𝐧𝐠 𝐨𝐧 𝐞𝐱𝐩𝐞𝐫𝐭-𝐝𝐞𝐟𝐢𝐧𝐞𝐝 𝐠𝐫𝐨𝐮𝐩𝐢𝐧𝐠𝐬! 🔥 #CurriculumLearning
March 24, 2025 at 3:09 PM
By capturing semantic relationships between goals, 🧭MAGELLAN enables efficient 𝐋𝐏 𝐞𝐬𝐭𝐢𝐦𝐚𝐭𝐢𝐨𝐧 & adaptive goal prioritization—all 𝐰𝐢𝐭𝐡𝐨𝐮𝐭 𝐫𝐞𝐥𝐲𝐢𝐧𝐠 𝐨𝐧 𝐞𝐱𝐩𝐞𝐫𝐭-𝐝𝐞𝐟𝐢𝐧𝐞𝐝 𝐠𝐫𝐨𝐮𝐩𝐢𝐧𝐠𝐬! 🔥 #CurriculumLearning
Our LLM agent uses 🧭MAGELLAN to estimate past & current competence, computing 𝐚𝐛𝐬𝐨𝐥𝐮𝐭𝐞 𝐥𝐞𝐚𝐫𝐧𝐢𝐧𝐠 𝐩𝐫𝐨𝐠𝐫𝐞𝐬𝐬 (𝐀𝐋𝐏) for each goal. The agent then selects goals that maximize ALP, learning efficiently via online RL. 🚀 #ReinforcementLearning #LLM
March 24, 2025 at 3:09 PM
Our LLM agent uses 🧭MAGELLAN to estimate past & current competence, computing 𝐚𝐛𝐬𝐨𝐥𝐮𝐭𝐞 𝐥𝐞𝐚𝐫𝐧𝐢𝐧𝐠 𝐩𝐫𝐨𝐠𝐫𝐞𝐬𝐬 (𝐀𝐋𝐏) for each goal. The agent then selects goals that maximize ALP, learning efficiently via online RL. 🚀 #ReinforcementLearning #LLM
🌍 Humans thrive in open-ended exploration, but AI struggles with infinite goal spaces. Learning progress (LP) helps, but scaling it is tough! 🧭MAGELLAN tackles this by efficiently generalising LP to goals not practice, allowing the agent to navigate large and complex domains🚢
March 24, 2025 at 3:09 PM
🌍 Humans thrive in open-ended exploration, but AI struggles with infinite goal spaces. Learning progress (LP) helps, but scaling it is tough! 🧭MAGELLAN tackles this by efficiently generalising LP to goals not practice, allowing the agent to navigate large and complex domains🚢