Gone are the days when all smart devices were required an internet uplink. The WiFi-enabled IoT fad, while still upon us (no, my coffee scale doesn’t need to be on …read more

The Inefficiency of Static Chain-of-Thought Reasoning in LRMs Recent LRMs achieve top performance by using detailed CoT reasoning to solve complex tasks. However, many simple tasks they handle could be solved by smaller models with fewer tokens, making such elaborate reasoning unnecessary. This echoes human thinking, where we use fast, intuitive responses for easy problems and slower, analytical thinking for complex ones. While LRMs mimic slow, logical reasoning, they generate significantly longer outputs, thereby increasing computational cost. Current methods for reducing reasoning steps lack flexibility, limiting models to a single fixed reasoning style. There is a growing need for adaptive reasoning that adjusts effort according to task difficulty.  Limitations of Existing Training-Based and Training-Free Approaches Recent research on improving reasoning efficiency in LRMs can be categorized into two main areas: training-based and training-free methods. Training strategies often use reinforcement learning or fine-tuning to limit token usage or adjust reasoning depth, but they tend to follow fixed patterns without flexibility. Training-free approaches utilize prompt engineering or pattern detection to shorten outputs during inference; however, they also lack adaptability. More recent work focuses on variable-length reasoning, where models adjust reasoning depth based on task complexity. Others study “overthinking,” where models over-reason unnecessarily. However, few methods enable dynamic switching between quick and thorough reasoning—something this paper addresses directly.  Introducing OThink-R1: Dynamic Fast/Slow Reasoning Framework Researchers from Zhejiang University and OPPO have developed OThink-R1, a new approach that enables LRMs to switch between fast and slow thinking smartly, much like humans do. By analyzing reasoning patterns, they identified which steps are essential and which are redundant. With help from another model acting as a judge, they trained LRMs to adapt their reasoning style based on task complexity. Their method reduces unnecessary reasoning by over 23% without losing accuracy. Using a loss function and fine-tuned datasets, OThink-R1 outperforms previous models in both efficiency and performance on various math and question-answering tasks.  System Architecture: Reasoning Pruning and Dual-Reference Optimization The OThink-R1 framework helps LRMs dynamically switch between fast and slow thinking. First, it identifies when LRMs include unnecessary reasoning, like overexplaining or double-checking, versus when detailed steps are truly essential. Using this, it builds a curated training dataset by pruning redundant reasoning and retaining valuable logic. Then, during fine-tuning, a special loss function balances both reasoning styles. This dual-reference loss compares the model’s outputs with both fast and slow thinking variants, encouraging flexibility. As a result, OThink-R1 can adaptively choose the most efficient reasoning path for each problem while preserving accuracy and logical depth.  Empirical Evaluation and Comparative Performance The OThink-R1 model was tested on simpler QA and math tasks to evaluate its ability to switch between fast and slow reasoning. Using datasets like OpenBookQA, CommonsenseQA, ASDIV, and GSM8K, the model demonstrated strong performance, generating fewer tokens while maintaining or improving accuracy. Compared to baselines such as NoThinking and DualFormer, OThink-R1 demonstrated a better balance between efficiency and effectiveness. Ablation studies confirmed the importance of pruning, KL constraints, and LLM-Judge in achieving optimal results. A case study illustrated that unnecessary reasoning can lead to overthinking and reduced accuracy, highlighting OThink-R1’s strength in adaptive reasoning.  Conclusion: Towards Scalable and Efficient Hybrid Reasoning Systems In conclusion, OThink-R1 is a large reasoning model that adaptively switches between fast and slow thinking modes to improve both efficiency and performance. It addresses the issue of unnecessarily complex reasoning in large models by analyzing and classifying reasoning steps as either essential or redundant. By pruning the redundant ones while maintaining logical accuracy, OThink-R1 reduces unnecessary computation. It also introduces a dual-reference KL-divergence loss to strengthen hybrid reasoning. Tested on math and QA tasks, it cuts down reasoning redundancy by 23% without sacrificing accuracy, showing promise for building more adaptive, scalable, and efficient AI reasoning systems in the future.  Check out the Paper and GitHub Page. All credit for this research goes to the researchers of this project. Also, feel free to follow us on Twitter and don’t forget to join our 100k+ ML SubReddit and Subscribe to our Newsletter. Sana HassanSana Hassan, a consulting intern at Marktechpost and dual-degree student at IIT Madras, is passionate about applying technology and AI to address real-world challenges. With a keen interest in solving practical problems, he brings a fresh perspective to the intersection of AI and real-life solutions.Sana Hassanhttps://www.marktechpost.com/author/sana-hassan/Building AI-Powered Applications Using the Plan → Files → Code Workflow in TinyDevSana Hassanhttps://www.marktechpost.com/author/sana-hassan/MemOS: A Memory-Centric Operating System for Evolving and Adaptive Large Language ModelsSana Hassanhttps://www.marktechpost.com/author/sana-hassan/Google AI Unveils a Hybrid AI-Physics Model for Accurate Regional Climate Risk Forecasts with Better Uncertainty AssessmentSana Hassanhttps://www.marktechpost.com/author/sana-hassan/Run Multiple AI Coding Agents in Parallel with Container-Use from Dagger

Reasoning tasks are a fundamental aspect of artificial intelligence, encompassing areas like commonsense understanding, mathematical problem-solving, and symbolic reasoning. These tasks often involve multiple steps of logical inference, which large language models (LLMs) attempt to mimic through structured approaches such as chain-of-thought (CoT) prompting. However, as LLMs grow in size and complexity, they tend to produce longer outputs across all tasks, regardless of difficulty, leading to significant inefficiencies. The field has been striving to balance the depth of reasoning with computational cost while also ensuring that models can adapt their reasoning strategies to meet the unique needs of each problem. A key issue with current reasoning models is the inability to tailor the reasoning process to different task complexities. Most models, including well-known ones like OpenAI’s o1 and DeepSeek-R1, apply a uniform strategy—typically relying on Long CoT across all tasks. This causes the “overthinking” problem, where models generate unnecessarily verbose explanations for simpler tasks. Not only does this waste resources, but it also degrades accuracy, as excessive reasoning can introduce irrelevant information. Approaches such as prompt-guided generation or token budget estimation have attempted to mitigate this issue. Still, these methods are limited by their dependence on predefined assumptions, which are not always reliable for diverse tasks. Attempts to address these issues include methods like GRPO (Group Relative Policy Optimization), length-penalty mechanisms, and rule-based prompt controls. While GRPO enables models to learn different reasoning strategies by rewarding correct answers, it leads to a “format collapse,” where models increasingly rely on Long CoT, crowding out more efficient formats, such as Short CoT or Direct Answer. Length-penalty techniques, such as those applied in methods like THINKPRUNE, control output length during training or inference, but often at the cost of reduced accuracy, especially in complex problem-solving tasks. These solutions struggle to achieve a consistent trade-off between reasoning effectiveness and efficiency, highlighting the need for an adaptive approach. A team of researchers from Fudan University and Ohio State University introduced the Adaptive Reasoning Model (ARM), which dynamically adjusts reasoning formats based on task difficulty. ARM supports four distinct reasoning styles: Direct Answer for simple tasks, Short CoT for concise reasoning, Code for structured problem-solving, and Long CoT for deep multi-step reasoning. It operates in an Adaptive Mode by default, automatically selecting the appropriate format, and also provides Instruction-Guided and Consensus-Guided Modes for explicit control or aggregation across formats. The key innovation lies in its training process, which utilizes Ada-GRPO, an extension of GRPO that introduces a format diversity reward mechanism. This prevents the dominance of Long CoT and ensures that ARM continues to explore and use simpler reasoning formats when appropriate. The ARM methodology is built on a two-stage framework. First, the model undergoes Supervised Fine-Tuning (SFT) with 10.8K questions, each annotated across four reasoning formats, sourced from datasets like AQuA-Rat and generated with tools such as GPT-4o and DeepSeek-R1. This stage teaches the model the structure of each reasoning format but does not instill adaptiveness. The second stage applies Ada-GRPO, where the model receives scaled rewards for using less frequent formats, such as Direct Answer or Short CoT. A decaying factor ensures that this reward gradually shifts back to accuracy as training progresses, preventing long-term bias toward inefficient exploration. This structure enables ARM to avoid format collapse and dynamically match reasoning strategies to task difficulty, achieving a balance of efficiency and performance. ARM demonstrated impressive results across various benchmarks, including commonsense, mathematical, and symbolic reasoning tasks. It reduced token usage by an average of 30%, with reductions as high as 70% for simpler tasks, compared to models relying solely on Long CoT. ARM achieved a 2x training speedup over GRPO-based models, accelerating model development without sacrificing accuracy. For example, ARM-7B achieved 75.9% accuracy on the challenging AIME’25 task while using 32.5% fewer tokens. ARM-14B achieved 85.6% accuracy on OpenBookQA and 86.4% accuracy on the MATH dataset, with a token usage reduction of over 30% compared to Qwen2.5SFT+GRPO models. These numbers demonstrate ARM’s ability to maintain competitive performance while delivering significant efficiency gains. Overall, the Adaptive Reasoning Model addresses the persistent inefficiency of reasoning models by enabling the adaptive selection of reasoning formats based on task difficulty. The introduction of Ada-GRPO and the multi-format training framework ensures that models no longer waste resources on overthinking. Instead, ARM provides a flexible and practical solution for balancing accuracy and computational cost in reasoning tasks, making it a promising approach for scalable and efficient large language models. Check out the Paper, Models on Hugging Face and Project Page. All credit for this research goes to the researchers of this project. Also, feel free to follow us on Twitter and don’t forget to join our 95k+ ML SubReddit and Subscribe to our Newsletter. NikhilNikhil is an intern consultant at Marktechpost. He is pursuing an integrated dual degree in Materials at the Indian Institute of Technology, Kharagpur. Nikhil is an AI/ML enthusiast who is always researching applications in fields like biomaterials and biomedical science. With a strong background in Material Science, he is exploring new advancements and creating opportunities to contribute.Nikhilhttps://www.marktechpost.com/author/nikhil0980/This AI Paper Introduces WEB-SHEPHERD: A Process Reward Model for Web Agents with 40K Dataset and 10× Cost EfficiencyNikhilhttps://www.marktechpost.com/author/nikhil0980/This AI Paper Introduces MMaDA: A Unified Multimodal Diffusion Model for Textual Reasoning, Visual Understanding, and Image GenerationNikhilhttps://www.marktechpost.com/author/nikhil0980/This AI Paper Introduces Differentiable MCMC Layers: A New AI Framework for Learning with Inexact Combinatorial Solvers in Neural NetworksNikhilhttps://www.marktechpost.com/author/nikhil0980/This AI Paper Introduces GRIT: A Method for Teaching MLLMs to Reason with Images by Interleaving Text and Visual Grounding

In a move Democrats warned would have disastrous consequences for the economy, the environment, and public health, the Republican-led Senate Thursday voted to block California’s electric-vehicle mandates, revoking the state’s right to implement the nation’s toughest emissions standards.    Republicans used the Congressional Review Act, or CRA, to overturn California’s long-standing authority under the Clean Air Act to request waivers from the Environmental Protection Agency to pass emissions standards stricter than federal rules and protect residents from dangerous air pollution. The move affects 17 other states and Washington, D.C., which have voluntarily adopted one or more of California’s stricter standards.  The CRA allows Congress to quickly rescind a rule within a limited time after it’s issued by a federal agency, allowing a simple majority vote rather than the 60 votes needed to advance legislation under the filibuster rule.  An aerial view of traffic on a smoggy day in Los Angeles in January 1985. [Photo: Ernst Haas/Getty Images] But both the Senate parliamentarian, the chamber’s official nonpartisan adviser, and the Government Accountability Office, the nonpartisan congressional referee, said the waivers are not rules and so are not subject to the Congressional Review Act. In defying the Senate parliamentarian, Democrats charged, the vote endangers not just the health of children and the climate but also decades of legal precedent and the integrity of the Senate itself. “Today, the Senate has done something unprecedented,” said Sen. Sheldon Whitehouse of Rhode Island late Wednesday night, after he and his Democratic colleagues spent the past several days urging Republicans to respect not just California’s authority under the law, but also Senate rules.  “Our actions and the ones that will follow from the procedural steps taken here today, over the next day or so will change the Clean Air Act, will change the Congressional Review Act, will change the rules of the Senate, and will do so by overruling the parliamentarian and breaking the filibuster—in effect, going nuclear,” Whitehouse said, referring to attempts to subvert the filibuster. “This isn’t just about California’s climate policies, and this isn’t just about the scope of the Congressional Review Act, and this isn’t just about eliminating the legislative filibuster,” said California Sen. Alex Padilla on the Senate floor Tuesday. The Trump administration’s EPA submitted California’s waivers for review by Congress “with full knowledge that they are not actually rules” subject to the CRA, Padilla said, opening the door for any agency to ask Congress to revoke regulations a new administration doesn’t like.  By mid-afternoon Thursday, Republicans moved to overturn California’s waivers through a procedural maneuver—giving the Senate the authority to determine what constitutes a rule for fast-track voting. They overturned waivers behind California’s rules to reduce tailpipe emissions from passenger vehicles and trucks, those regulating medium- and heavy-duty trucks, and the rule for heavy-duty smog-producing diesel and gas trucks. Senate Majority Leader John Thune (R-SD) mocked Democrats’ objections to using the CRA, saying they were “throwing a tantrum over a supposed procedural problem.”Thune insisted that having a waiver submitted to Congress “is all that Congress has ever needed to decide to consider something under the Congressional Review Act.” He called the GAO’s ruling that the waiver is not a rule “an extraordinary deviation from precedent,” saying it was the first time the office “has decided to insert itself into the process and affirmatively declare that an agency rule submitted to Congress as a rule is not a rule.”  Despite Thune’s claim, since the CRA was passed in 1996 the GAO has offered 26 legal opinions about whether an agency action was a rule in response to inquiries from members of Congress. And EPA never submitted California Clean Air Act waivers to Congress before the Trump administration, Padilla and his Democratic colleagues say. They contend that Republicans chose this route because they don’t have the votes to withdraw the waivers through legislation. “The CRA has never been used to go after emission waivers like the ones in question today,” Senate Minority Leader Chuck Schumer of New York said on the floor Tuesday. “The waiver is so important to the health of our country, and particularly to our children; to go nuclear on something as significant as this and to do the bidding of the fossil fuel industry is outrageous.” The first waiver was granted to California on July 11, 1968, Whitehouse told his colleagues in a last-ditch effort to change their minds late Wednesday night. Waivers have either been granted or amended or modified repeatedly since then, he said. “The score on whether the California clean air rule is treated by EPA as a waiver or a rule? It’s 131 to zero.” The use of the Congressional Review Act resolution is inconsistent with past precedent and violates the plain language of the act itself, said John Swanton, a spokesperson for California’s Air Resources Board, which regulates emissions.  “The vote does not change CARB’s authority,” Swanton said, adding that the agency will continue its mission to protect the public health of Californians impacted by harmful air pollution. Ten million Californians live in areas that are under distinct, elevated threats from air pollution, said Adam Schiff, California’s junior senator. That has led to higher rates of respiratory issues like asthma and chronic lung disease, and increased the risk of heart disease, cancer, chronic headaches, and immune system issues, he said.  Sen. Adam Schiff (D-CA) speaks about the importance of the Clean Air Act in California during a Senate meeting on May 8. [Image: U.S. Senate floor webcast] “And that is multiplied by us living now on the front lines of the climate crisis. We have devastating and year-round fire dangers that put millions of other pollutants into our air,” Schiff said. “We need, deserve, and reserve the right as Californians to do something about our air.” Yet earlier this month, House Republicans, joined by 35 Democrats, including two from California, voted to rescind the waivers, sending the issue to the Senate. A “Compelling and Extraordinary” Need California’s legal authority to implement stricter air quality standards than federal rules comes from having already implemented its own tailpipe-emission regulations before Congress passed national standards in 1967. California officials developed the regulations to deal with the “compelling and extraordinary” air-pollution problems caused by the Golden State’s unique geography, climate, and abundance of people and vehicles. Recognizing these unique conditions, Congress gave California the authority to ask the Environmental Protection Agency for a waiver from rules barring states from passing air and climate pollution rules that are more protective than federal rules.  Only one waiver was denied, an action that was quickly reversed, according to CARB. And though the Trump administration in 2019 withdrew a waiver, a move legal scholars say has no basis in the law, the Biden administration restored the state’s authority to set its own vehicle-emission standards within a few years. Republicans argued that California’s rules amount to de facto national standards, given the state’s size and the fact that other states have signed on.  But California can’t force its emission standards on other states, Padilla said. “Yes, over a dozen other states have voluntarily followed in California’s footsteps, not because they were forced to, but because they chose to, in order to protect their constituents, their residents, and protect our planet.” California’s standards also represent ambitious but achievable steps to cut carbon emissions and fight the climate crisis, Padilla said. “Transportation is the single largest contributor to greenhouse gas emissions, and California has been proud to set the example for other states who may choose to follow suit.” Sen. Alex Padilla (D-CA) told his Republican colleagues late Wednesday night why his state’s unique geography and climate create particularly hazardous air-quality problems. [Image: U.S. Senate floor webcast] Padilla, who grew up in California’s chronically polluted San Fernando Valley, recalled being sent home from grade school “on a pretty regular basis” when throat-burning smog settled over the valley. “It appears that Republicans want to overturn half a century of precedent in order to undermine California’s ability to protect the health of our residents,” Padilla said. “Republicans seem to be putting the wealth of the big oil industry over the health of our constituents.” “For Their Fossil Fuel Donors” Rhode Island’s Whitehouse, who has long schooled his colleagues on the perils of carbon pollution, took to the floor Tuesday to school them on the Congressional Review Act. Under the American legal system, administrative agencies can make rules through “a very robust process” that follows the Administrative Procedure Act, Whitehouse said. A rule could be contested in court, but years ago Congress decided there also could be a period of review when congressional members could reject the rule.  And for all the decades since the CRA was passed, he said, it’s been used to address rules under the APA within the specified 60 days.Other states, including Rhode Island, follow California’s emissions standards because it’s good for public health to have clean air, Whitehouse said. “Efficient cars may mean lower cost for consumers, but those lower costs for consumers are lower sales for the fossil fuel industry.” Whitehouse told his colleagues they had legitimate pathways to change laws they didn’t like. They could pass a joint resolution or a simple Senate resolution. But those approaches would require 60 votes to end debate. “They don’t want to do that,” he said. “They want to ram this thing through for their fossil fuel donors.” Republicans, by contrast, argued they had the authority to protect consumers from what they call California’s “electric vehicle mandate,” which they say would endanger consumers, the economy, and the nation’s energy supply. “And our already shaky electric grid would quickly face huge new burdens from the surge in new electric vehicles,” argued Thune.  Congress had approved $5 billion to build electric vehicle charging infrastructure across the country, but the Trump administration withheld that funding, triggering a lawsuit from a coalition of attorneys to reverse what they said was a clearly illegal action. Republicans’ attacks on electric vehicles could disrupt a burgeoning industry built around the transition to renewable energy. “The repeal of these waivers will dramatically destabilize the regulatory landscape at a time when industry needs certainty to invest in the future and compete on a global scale,” said Jamie Hall, policy director for EV Realty, which develops EV-charging hubs. Thune also argued that California’s waiver rules are an improper expansion of a limited Clean Air Act authority, echoing an argument in Project 2025, a policy blueprint for the second Trump administration produced by the conservative Heritage Foundation, which has long battled efforts to combat climate change. In a chapter on transportation asserts, Project 2025 claims that California has no valid basis under the Clean Air Act to claim an extraordinary or unique air quality impact from carbon dioxide emissions. Its recommendation? “Revoke the special waiver granted to California by the Biden administration.” On Wednesday, a clearly frustrated Whitehouse argued that Republicans were helping the fossil fuel industry create a shortcut for itself so it can sell more gasoline and ignore all the states that joined California to demand cleaner air for their constituents. “The fossil fuel industry essentially runs the Republican Party right now,” he said. Last year, the oil and gas industry spent more than $153 million on lobbying, led by the American Fuel and Petrochemical Manufacturers, which spent $27.6 million to influence Congress on bills including those designed to repeal vehicle-emission standards. The trade group also donated $178,750 to congressional candidates, 96% of which went to Republicans.  The American Petroleum Institute, the largest U.S. oil and gas industry trade association, spent $6.25 million on lobbying last year to influence some of the same bills. Of nearly $400,000 donated to congressional candidates last year, 78% went to Republicans.  Ninety-five percent of the $21,000 the Heritage Foundation donated to congressional candidates last year went to Republicans. “We Believe That You Can Do It” The week before Donald Trump returned to office, the American Petroleum Institute held its biggest annual meeting in Washington, D.C. API promoted the event as an opportunity to urge the incoming Trump administration and Congress to “seize the American energy opportunity” by advancing commonsense energy policies. Thune joined API Chief Executive Mike Sommers onstage, where they reminisced about starting their careers in adjacent offices in the same congressional office building 30 years ago.  “It is a huge opportunity, having an administration that actually is pro-energy development working with the Congress,” Thune told his old friend. “We want to be supportive in any way that we can in ensuring that the president and his team have success in making America energy dominant.” Sommers suggested that one of the “big, powerful tools” Congress can use when one party controls both chambers is the Congressional Review Act, which he said offers fast-track authority to reverse “midnight regulations” passed by the Biden administration. Thune said he wouldn’t be able to use the CRA for one of California’s tailpipe emissions standards because it doesn’t fit within the required time window. But he was arguing with the parliamentarian and others, he said, “about the whole California waiver issue and how to reverse that because that was such a radical regulatory overreach.”Both California’s Clean Cars and Clean Trucks rules require an increasing percentage of vehicles sold in the state to be zero-emissions by 2035, with the cars rule, the so-called “EV mandate,” requiring that 100% of passenger cars and trucks be zero emissions by that date. “What California did was completely radical,” Sommers said at the meeting. “The fact that 17 other states who’ve waived into this are going to be subject to it could completely change the vehicle market.” “So we would highly encourage you to look at that as an option for the CRA,” Sommers told Thune. “And we believe that you can do it.” Thune assured Sommers that his committee chairs and team were looking at ways to fit repeal of California’s waivers “within the parameters of a CRA action” to fix what they saw as a shared problem. The oil and gas industry appreciated the efforts of Thune; John Barrasso of Wyoming, the Senate Majority Whip; and West Virginia Sen. Shelley Moore Capito, who pledged to overturn California’s clean cars rule and introduced the measure to do so last month.  “Today, the United States Senate delivered a victory for American consumers, manufacturers, and U.S. energy security by voting to overturn the prior administration’s EPA rule authorizing California’s gas car ban and preventing its spread across our country,” said the American Petroleum Institute and the American Fuel and Petrochemical Manufacturers in a joint statement. “We cannot thank Senators John Barrasso, Shelley Moore Capito, and Leader John Thune enough for their leadership on this important issue.” Back on the Senate floor, Democrats warned their Republican colleagues that they may live to regret their decision to override the parliamentarian and flout legislative rules. “It won’t be long before Democrats are once again in the driver’s seat here, in the majority once again,” Padilla said. When that happens, he warned, every agency action that Democrats don’t like, whether it’s a rule or not, and no matter how much time has passed, would be fair game with this new precedent.  “I suggest that we all think long and hard and be very careful about this,” he implored, in vain. “I would urge my colleagues, all my colleagues, to join me, not just in defending California’s rights to protect the health of our residents, not just in combating the existential threat of climate change, but in maintaining order in this chamber.” This article originally appeared on Inside Climate News. It is republished with permission. Sign up for their newsletter here.