Ursula Spoj

[Mario Nawfal @MarioNawfal]

INVESTMENT: More than $500M was lost to memecoin rug pulls & scams in 2024 alone. Just a few days ago Bybit suffered a major breach totaling in over $1.4B loss.

This is becoming a negative trend and someone has decided to do something about it.

@veritas_web3 offers a set of tools that let anyone audit tokens and any project's smart contracts 90% faster and at just 10% of the cost compared to traditional solutions.

Why is this significant?

Because traditional audits can cost tens to hundreds of thousands of dollars. An indie Web3 team can't afford this unless it sells a kidney.

Veritas' suite of tools includes automatic round-the-clock audits with minimal manual intervention, an AI debugger, and insurance against exploits.

It is like obtaining a full smart contract audit and security service, at the price of a can of beer.

Their token is getting listed on February 26, 12 pm UTC, on MEXC and Uniswap.

Know more on https://t.co/877Z4fyMt2.

Disclaimer: We are investors in this project and receive benefits from our involvement. This content is for promotional purposes only and is not financial or investment advice. Cryptocurrency carries significant risk, and you should always do your own due diligence before making any decision.

[Anndy Lian @anndylian]

Look at on chain data. Look for tokens with 10,000 holders and at least 500 active on X. Those are potential gems.

[EurexaLabs @EurexaLabs]

To meet the first criterion, researchers follow Zero-shot CoT and first convert the input data example into a prompt with a simple template “Q: [X]. A: [T]”. Specifically, the input slot [X] contains the input problem statement and a hand-crafted instruction is specified in the input slot [T] to trigger LLMs to generate a reasoning process that includes a plan and steps to complete the plan. In Zero-shot-CoT, the instruction in the input slot [T] includes the trigger instruction Let’s think step by step”. Their Zero-shot PS prompting method instead includes the instructions “devise a plan” and “carry out the plan” . Thus, the prompt would be “Q: [X]. A: Let’s first understand the problem and devise a plan to solve the problem. Then, let’s carry out the plan and solve the problem step by step.”We then pass the above prompt to the LLM which subsequently outputs a reasoning process. In accordance with Zero-shot-CoT, researchers' method uses the greedy decoding strategy (1 output chain) for generating output by default. 3/6

[EurexaLabs @EurexaLabs]

Experimental Results-Main Results
Arithmetic Reasoning. Table 2 reports the accuracy comparison of researchers‘ method and existing zeroshot and few-shot methods on the arithmetic reasoning datasets. In the zero-shot setting, their PS+ prompting (i.e., PS prompting with more detailed instructions) consistently outperforms Zero-shotCoT across all arithmetic reasoning datasets by a large margin. Specifically, PS+ prompting improves the accuracy over Zero-shot CoT by at least 5% for all datasets except GSM8K which sees a 2.9% improvement. The exception could be due to GSM8K being a more challenging dataset from the linguistics complexity aspect. 1/5

[EurexaLabs @EurexaLabs]

In the final section of our project, we will focus on introducing 🔨 Joint Mining Mechanism. Behavioral economics research shows that people’s decisions are often influenced by subconscious biases, especially when assessing risks and rewards. In our framework, by providing long-term incentives for early investments, we can effectively motivate participants to continually invest resources and innovate technologically. This incentive mechanism not only enhances the efficiency of model development but also ensures that the interests of both parties in the collaboration are maximized. 1/5

[EurexaLabs @EurexaLabs]

The results suggest that adding more detailed instructions to the prompt can effectively elicit higher-quality reasoning steps from LLMs. Compared with the few-shot methods, Manual CoT and Auto-CoT, PS+ prompting yields an average accuracy (76.7%) slightly lower than ManualCoT (77.6%) but higher than Auto-CoT (75.9%). While this is an unfair comparison, this result indicates that zero-shot prompting can outperform few shot CoT prompting, which hopefully will spark further development of new ways with a less manual effort to effectively elicit reasoning in LLMs. 3/5

[EurexaLabs @EurexaLabs]

Computing Power Providers - Computing Power Investors 🧮
In traditional computing power supply mechanisms, computing power providers mainly refer to GPU card providers. This role was previously more like a pure hardware supplier, but under our framework, computing power providers now have their own "operating system." This is similar to how IBM hardware once ran Microsoft’s operating system. We hope that in the AI development cycle, hardware and computing power providers can truly inject soul into their contributions. We have opened up this new model of computing power investment to help computing power providers transform short-term returns into long-term benefits. In traditional cooperation models, regardless of how innovative the model becomes or how many people use it, computing power providers rarely share in the benefits brought by the model. The joint mining mechanism effectively balances this inequality in the benefit cycle. This model is actually quite similar to some marriage relationships: women need to invest significant time and energy early on, sacrificing career opportunities to take on the responsibilities of childbirth and raising the next generation; while men often only manage to give back to the family through promotions and salary increases many years later. Marriage is, to some extent, a mapping of the joint mining model—we need to ensure that the party who contributes more in the early stages can fairly share in the long-term benefits. 2/5

[EurexaLabs @EurexaLabs]

Conclusion✍️
Rsearchers find that Zero-shot-CoT still suffers from three pitfalls: calculation errors, missingreasoning-step errors, and semantic understanding errors. To address these issues, they introduce plan-and-solve prompting strategies (PS and PS+ prompting). They are new zero-shot prompting methods that guide LLMs to devise a plan that divides the entire task into smaller subtasks and then carries out the subtasks according to the plan. Evaluation on ten datasets across three types of reasoning problems shows PS+ prompting outperforms the previous zero-shot baselines and performs on par with few-shot CoT prompting on multiple arithmetic reasoning datasets. 1/4

[EurexaLabs @EurexaLabs]

Our tweets will feature more about Robotics in the future to suitable our theme. Today, we are introducing the first project: HOVER: Versatile Neural Whole-Body Controller for Humanoid Robots. Humanoid whole-body control requires adapting to diverse tasks such as navigation, loco-manipulation, and tabletop manipulation, each demanding a different mode of control. For example, navigation relies on root velocity or position tracking, while tabletop manipulation prioritizes upperbody joint angle tracking. Existing approaches typically train individual policies tailored to a specific command space, limiting their transferability across modes. Therefore researchers present the key insight that full-body kinematic motion imitation can serve as a common abstraction for all these tasks and provide generalpurpose motor skills for learning multiple modes of whole-body control. Building on this, they propose HOVER (Humanoid Versatile Controller), a multi-mode policy distillation framework that consolidates diverse control modes into a unified policy. HOVER enables seamless transitions between control modes while preserving the distinct advantages of each, offering a robust and scalable solution for humanoid control across a wide range of modes. By eliminating the need for policy retraining for each control mode, this approach improves efficiency and flexibility for future humanoid applications. 1/6

[EurexaLabs @EurexaLabs]

Commmonsense Reasoning. Table 3 shows the results on commonsense reasoning datasets: CommonsenseQA and StrategyQA. Researchers only include their better zero-shot PS+ prompting strategy in this comparison. Zero-shot PoT is excluded as it does not work on this problem. While PS+ prompting underperforms Few-Shot-CoT(Manual) on this problem, it consistently outperforms Zero-shot CoT on CommonsenseQA (71.9% vs. 65.2%) and StrategyQA (65.4% vs. 63.8%) datasets. 4/5

[EurexaLabs @EurexaLabs]

They have developed a range of individual sports environments, such as golf, javelin throw, high jump, long jump, and hurdles. Additionally, they have set up competitive sports, including both 1v1 and 2v2 games like table tennis, tennis, fencing, boxing, soccer, and basketball. Their research indicates that by combining robust motion priors with straightforward rewards, the humanoids can exhibit human-like behavior in these sports. By offering a unified sports benchmark and a baseline implementation for state and reward designs, they hope that SMPLOlympics will assist the control and animation communities in achieving more human-like and performant behaviors. 2/6

[EurexaLabs @EurexaLabs]

A. Comparison with Specialists
Comparison with Specialists of Prior Work’s Control Mode. To address Q1 (Can HOVER as a generalist policy outperform policies trained for a specific command configuration?), researchers compare the performance of the same HOVER policy across different control modes against corresponding specialist policies. For example, the performance of HOVER under ExBody mode is evaluated with a fixed mask to match ExBody mode across the entire dataset Qˆ. HOVER consistently demonstrates superior generalization. In every command mode, HOVER outperforms prior work specialist controllers in at least 7 out of the 12 metrics. This consistent advantage across various control modes underscores the versatility of HOVER.
Furthermore, this means that even when focusing on a single control mode without considering multi-mode versatility, distilling from an oracle policy still surpasses RL-trained specialists. Comparison with Other Specialists of Other Useful Control Mode. In addition to the aforementioned baselines, they also evaluate four additional modes: left-hand mode, righthand mode, two-hand mode, and head mode. They train four RL specialists to track these modes individually. The results show that HOVER consistently outperforms specialists in terms of tracking metrics that are trained for specific command configurations. 4/7

[EurexaLabs @EurexaLabs]

Which of the following robotics companies do you like the most? 🤖

[EurexaLabs @EurexaLabs]

Introduction
Competitive sports, just like in human society, provide a standardized way to measure the performance of learning algorithms and create realistic human-like behavior. However, previous attempts to bring individual sports into physics simulations have been scattered and inconsistent, using different humanoids, simulators, and learning algorithms, which has made it hard to compare results. Plus, these specially designed humanoids can be tricky to work with when it comes to getting compatible motion data, often requiring retargeting to fit each humanoid. Creating a collection of simulated sports environments that all use the same humanoid design and training process is a tough nut to crack, requiring deep knowledge of humanoid design, reinforcement learning (RL), and physics simulation. That's why previous benchmarks and simulated environments have mainly focused on basic locomotion tasks for humanoids, like walking, standing up, and navigating different terrains. While these tasks are important benchmarks, they don't cover the range of behaviors and strategies needed to mimic real-world human activities. 3/6

[EurexaLabs @EurexaLabs]

Starting today, we are introducing another robotics project - the SMPLOlympics.
The team is introducing a fascinating new project called SMPLOlympics. It's a series of physically simulated environments where humanoid robots can engage in various Olympic sports. These sports simulations serve as an excellent platform for evaluating and refining learning algorithms, given the diversity and physical demands of athletic activities. Furthermore, humans have been participating in these sports for many years, resulting in a wealth of existing knowledge about effective strategies. To capitalize on this human expertise from videos and motion capture, the team designed their humanoid robots to be compatible with the widely-used SMPL and SMPL-X human models in the vision and graphics community. 1/6

[VERITAS PROTOCOL @veritas_web3]

$VPT is trading LIVE!

Chain: @base
DEX: @Uniswap
CEX: @MEXC_Official

CA: 0x00096697DC24bd10423690126D91546A20Ccb3F0

Beware of fake contracts, only take information from our official channels.

🌐 https://t.co/H9aYp9MnOc

💬 https://t.co/lN2Sr7O5WA

[Fans Musicales No oficiales @MusicalesNo]

Miley Cyrus le dedica unas emotivas palabras a Ariana Grande en su último vídeo: "Es una amiga de verdad"

[VERITAS PROTOCOL @veritas_web3]

🤝 VERITAS × MARIO NAWFAL

Veritas Protocol and @MarioNawfal join forces.

What this means:
🔹One of crypto's most influential voices backs Web3 security leader
🔹Access to @ibcgroupio's 2M+ follower network
🔹Strategic partnership connections

AI powered Web3 security just went mainstream.

[VERITAS PROTOCOL @veritas_web3]

🤝 VERITAS × https://t.co/V7CxKYsenE

@rwa_io selects Veritas Protocol for major security initiative: Full audit of Real World Asset sector.

Phase 1 (already in progress):
➡️ Smart contract Trust Score for 240+ projects listed on https://t.co/V7CxKYsenE Insights
➡️ Smart contract verification for upcoming sales on https://t.co/V7CxKYsenE Launchpad

Phase 2 (in the pipeline):
➡️ Risk assessment system for investable collections (index funds)
➡️ Security standards for RWA tokenization

Building the foundation for secure asset tokenization with @rwa_io.