It is amazing how the cultural zeitgeist can influence the public fear perception One example is Mushrooms This has long been ‘a thing’ to some extent for reasons I won’t go into but anyone myco-curious would know Yet a cultural incident that sweeps the cultural zeitgeist can re-influence the perception There has been one mushroom in particular that has shaped my journey physically, spiritually and mentally and the bridge for me was the urgent need for natural pain relief as I decided to cease taking painkillers I had relied upon since my teens on a regular basis Pain has been the pain point for me that was a door from the world and life I felt forced into, opening out into the world and life I surrender into As the path unfolds, this Mushroom appears just as I need to prepare another oil for topical use 🍄 Collecting these mushrooms, most onlookers (aside from one knowing family) gazed with expressions that could be read as easily as a book 📖 evidently suspicious we were off to either poison someone or “get high” (In hindsight I could have told them “well of course, they are red so they are great for period pain” but this is only occurring to me now, as I write this… one family passing by surprised us, “Amanita Muscaria” one member called out knowingly) Of course, the perception of poison in our world and culture is incredibly skewed; just look at the products ands brands considered safe for household or bodily use, the pain killers I once used included It is time we consider where we place our trust, in my view, and return that to the natural world and away from some of these brands that have gaslit us into believing their highly marketed definitions of the word ‘safe’ And the “get high” perception that I nowadays see as the cultural norm, particularly in the party or social culture outside of healing spaces also needs evolving Rather, every medicine of the natural world, whether that is of a plant origin or fungi, best be revered and entered into sacred relation with through a silent dance that is born of mutual respect *before* sacred connection (just as any intimacy is best honoured) and so the idea of the outcome is not the focus; the relationship is Through the interrelation can come new understandings of self and the body system Through this inner knowing one can dance with the medicine at the subtle level without needing to achieve a state of feeling that is “higher” than the usual state, or a point I think of as zero point Knowing your own body system and what it’s zero point feels like and understanding when you are off balance and how medicines of the natural world can guide you towards your own sacred balance is, in my view, a part of how we can return to knowing our own sacred healing system And this is of course not limited to the medicine of plants and fungi, this would include water, earth, air, fire, sound, movement… and many other natural preparations of elements So we return to self as the ultimate medicine knowing that this is the most loving thing we can do for the world: Love self Listen within Know self ❤️ Heal the world #loveself #healtheworld #selflove #knowself #sacredbalance #mushroom #taboo #beyondfear

Verses/hooks in here from @Birdapres, @Buck65, Sixtoo, @Tachichiscotia, Fatt Matt, CEE!!!, Kunga 219, @Mokaonly_raps, @AWOLONE, @BigJoshMartinez, Kaboom, LXVNDR, Paul the Apostle, RC, King T, Jay O Smooth, Cheklove Shakil, Knowself, Al Turdbeast, Kotep Omegatron...

ALL IS ONE #sacredgeometry #divinemathematics #Oneness #knowself

Lotta people would be much less angry if their trauma was even recognized, let alone understood… #leggodebaggage #freeupandlive #Namasteinyalane #knowself

#knowself #allisONE

And they were made into the separate religions to keep us all divided… Sun- Christian Moon- Islam Stars- Judaism When it’s all actually the ONENESS of creation/life. All things are ONE living thing directed by the celestial movements. #knowself

This the only son the black delegation recognizes… King of Kings Moa Ambessa ze imnegede Yehuda Power of the Trinity HIM Haile Selassie I 💚💛❤️🖤 #LionSeasonBegins #HBD #sonlight #sunraise #knowself #AfricaUnite

When we come to realize that it’s actually jes US 😉 WE will make the world what it could be. The creation…one with the creator. We can bring forth light, or plunge into darkness…all our choice! 🌞 #sonlight #oneness #ALLISONE #knowself #everythingisarichmanstrick

Looka this you lot… I never cease to be amazed at the talent that humans possess. #tapintothesource #beyou #creationscreationcreatecreations #theartistandtheartareone #fromtheroottothefruit #creationisforever #lifeisforever #oneness #knowself #artislife #lifeiscreation

Spirit, soul, body is Mind, soul, body is Man, woman, child is Peace, love, harmony is Sun, moon, earth is 120, 120, 120degrees #360degreesofself #oneness #tetrahedron #atom #sonlight #sunraise #unity #community #lifeislove #loveisGod #Godiseverything #knowself #KemetNu

🤯 What if an AI agent could sense when it’s out of its depth—pause, reflect, or fetch just the right piece of knowledge before acting? That’s exactly what “KnowSelf”––a new paradigm of agentic knowledgeable self‑awareness––brings to Large Language Model (LLM) agents. Here are the key takeaways from a newly published research paper 1. KnowSelf trains agents to switch between fast thinking (act immediately), slow thinking (reflect first), and knowledgeable thinking (retrieve external rules) by inserting special tokens into their own trajectories. This mirrors human metacognition without flooding the prompt with extra data. 2. Less knowledge ≠ less performance! On the ALFWorld benchmark, an 8‑B Llama agent using KnowSelf hit 84.3 % average reward with only 15 % of steps augmented by knowledge, beating 100 %‑knowledge baselines like ExpeL (79.9 %) and even matching GPT‑4o‑level agents. On WebShop it scores 67.1 % with 17 % knowledge, edging out heavier fine‑tuned models. 3. Selective knowledge calls cut inference cost (fewer tokens, API hits) and improve generalisation: when trained on three household tasks, KnowSelf outperformed reflection‑only agents on unseen tasks such as Heat and Cool. And scaling tests show self‑awareness benefits appear once ~40 % of the training data teaches these modes—promising for small, edge‑deployed models. Why this all matters? - Imagine IDE co‑pilots that pull docs only when their confidence dips. - Household robots or e‑commerce bots can plan robustly without constant API queries. - Opens the door to architectures that learn when to reason versus when to recall. With all that said.... Experiments are limited to two simulated datasets and ≤8 B models; multimodal perceptions and real‑world latency remain open challenges—ripe territory for the community. Could the next wave of LLM agents manage their own “attention budget” better than we do? Full paper link in the comments. #LLMAgents #MachineLearning #AI #DataScience #AutonomousAgents

Can’t tell me my people ain’t from de ancients… 💚💛❤️🌍❤️🖤💚 #eyeinIaredeROOT #dogon #knowself #gombey #junkanoo #diaspora #diasporaconnect #carriedbeings #caribbean #africanheriitage #Africa #sunpeople #naturalmystic #higherheights #livity #Bermuda

One of my great great grandfathers, on my mamas side, Frank Pires..was a Cape Verdean #knowself #lineage #mystory

8). Agentic Knowledgeable Self-awareness KnowSelf is a new framework that introduces agentic knowledgeable self-awareness, enabling LLM agents to dynamically decide when to reflect or seek knowledge based on situational complexity, mimicking human cognition. arxiv.org/abs/2504.03553v1

7. KnowSelf Proposes “agentic knowledgeable self-awareness” for agent models, enabling dynamic self-regulation of knowledge use during planning and decision-making arxiv.org/abs/2504.03553 Code: github.com/zjunlp/KnowSelf

The freshest AI/ML research of the week Our top 10: ▪️ AI for Software Engineering ▪️ Inference-Time Scaling for Reward Modeling ▪️ Inference-Time Scaling for Complex Tasks ▪️ Open-Reasoner-Zero ▪️ SynWorld ▪️ Agent S2 ▪️ KnowSelf ▪️ ZClip ▪️ MegaScale-Infer ▪️ Scaling Laws in Scientific Discovery ▪️ Trustworthy GUI Agents ▪️ Harnessing the Reasoning Economy ▪️ Test-Time Scaling in LLMs ▪️ Efficient Inference for Large Reasoning Models ▪️ Efficient Reasoning for LRMs ▪️ GenPRM ▪️ Thinking Intervention ▪️ Recitation over Reasoning ▪️ JudgeLRM ▪️ Improved Visual-Spatial Reasoning ▪️ Interpreting Emergent Planning in RL ▪️ Scaling Visual SSL ▪️ Multi-Token Attention ▪️ VerifiAgent ▪️ Scaling Speech-Text Models ▪️ MegaMath ▪️ RLHF Data Scaling ▪️ Massive Activations ▪️ Instruction-Guided Parameter Generation (IGPG) ▪️ Efficient Model Selection via LLMs 🧵

Sure! Here's a step-by-step explanation of the "Agentic Knowledgeable Self-awareness" framework (KnowSelf) based on the provided visuals and context: 1. Concept Overview: KnowSelf enables agents to dynamically assess situations and adapt their behavior by switching between three modes of thinking: Fast Thinking: For simple tasks the agent can handle directly. Slow Thinking: For moderately challenging tasks requiring reflection. Knowledgeable Thinking: For complex tasks where external knowledge is needed. 2. Framework Steps: Step 1: Self-awareness Data Construction Agents explore their environment and log actions, observations, and reflections. They classify situations using a heuristic criterion into fast, slow, or knowledgeable thinking. Example: If an egg is dirty, the agent reflects on how to clean it or finds a clean one instead. Step 2: Self-awareness Learning The agent undergoes a two-stage training process: Supervised Fine-Tuning (SFT): Teaches initial self-awareness patterns. Reinforcement with RPO Loss: Optimizes long-term planning by refining decision-making strategies. Step 3: Self-awareness Inference During task execution, the agent identifies situations in real time and generates special tokens (e.g., <reflection>, <knowledge>). Example: When tasked to put two newspapers in a drawer, it reflects on errors and incorporates knowledge to prioritize correct actions. 3. Performance Highlights: KnowSelf was tested on ALFWorld tasks using Llama-8B and Gemma-2B models. It outperformed baseline methods by efficiently balancing reflection and knowledge use, achieving higher task success rates with minimal resource usage. Let me know if you'd like more details about any specific part! 😊

KnowSelf is a data-centric method for adaptive knowledge use in LLM agents. It avoids standard trajectory/feedback injection by training agents to self-regulate based on context. A heuristic marks control tokens in agent-generated traces, guiding a two-stage process that enables situational strategy switching. Outperforms baselines with reduced external supervision across planning tasks.

KnowSelf = promising paper here for agents 👇 Interesting integration of situational awareness self-awareness in agents. I also like the agentic situations triptych. Curious to see if the results would be the same on other tasks. Bravo to the team 👏

Tired of agents that mindlessly follow patterns? 🤖💡 We introduce "Agentic Knowledgeable Self-awareness", enabling agent to dynamically assess situations and strategically use resources! Paper: huggingface.co/papers/2504.0… Code: github.com/zjunlp/KnowSelf 🧠 Key Idea: Agents learn to recognize when they need to reflect, when to seek knowledge, and when to act directly. This avoids overfitting to planning patterns and reduces unnecessary knowledge usage. 🔍 How It Works: 1️⃣ Data-Driven Training: Special tokens mark fast, slow, and knowledgeable thinking. 2️⃣ Two-Stage Learning: Supervised fine-tuning RPO loss for robust self-awareness. 3️⃣ Inference: Agents generate tokens to reflect or query knowledge based on context. 📈 Results: Outperforms baselines with minimal knowledge usage! Breaks pattern overfitting and enhances generalization! Scales efficiently with model size and training data! 🌐 Future: KnowSelf paves the way for smarter, more efficient agents. Maybe in the future, we can train models to develop stronger agentic self-awareness through reinforcement learning powerful verifier engineering. #AI #LLMs #AgentPlanning #SelfAwareness #NLP #agent #KnowledgeAugmentation