Aligning large language models (LLMs) to human preferences typically relies on aggregating pooled feedback into a single reward model. However, this standard approach assumes that all labelers share the same underlying preferences, ignoring the fact that real-world labelers are highly heterogeneous and usually anonymous. Consequently, relying solely on binary choice data fundamentally distorts the learned policy, making the true population-average preference unidentifiable. To overcome this critical limitation, we demonstrate that augmenting preference datasets with a simple, secondary signal -- the user's response time -- can restore the identifiability of the population's average preference. By modeling each decision as a Drift-Diffusion Model (DDM), we introduce a novel, consistent estimator of heterogeneous preferences that successfully corrects the distortions of standard choice-only labels. We prove that our estimator asymptotically converges to the true average preference even in extreme cases where each anonymous labeler contributes only a single choice. Empirically, across both synthetic and real-world datasets, our method consistently outperforms standard baselines that otherwise fail and plateau at a bias floor. Because response times are essentially free to record and require zero user tracking or identification, our results bring promises and open up new opportunities for future data-collection pipelines to improve the social benefit without requiring user-level identifiers or repeated elicitations.

One of the most common machine learning setups is logistic regression. In many classification models, including neural networks, the final prediction is obtained by applying a logistic link function to a linear score. In binary logistic regression, the feedback can be either soft labels, corresponding to the true conditional probability of the data (as in distillation), or sampled hard labels (taking values $\pm 1$). We point out a fundamental problem that arises even in a particularly favorable setting, where the goal is to learn a noise-free soft target of the form $σ(\mathbf{x}^{\top}\mathbf{w}^{\star})$. In the over-constrained case (i.e. the number of samples $n$ exceeds the input dimension $d$) with examples $(\mathbf{x}_i,σ(\mathbf{x}_i^{\top}\mathbf{w}^{\star}))$, it is sufficient to recover $\mathbf{w}^{\star}$ and hence achieve the Bayes risk. However, we prove that when the examples are labeled by hard labels $y_i$ sampled from the same conditional distribution $σ(\mathbf{x}_i^{\top}\mathbf{w}^{\star})$ and $\mathbf{w}^{\star}$ is $s$-sparse, then rotation-invariant algorithms are provably suboptimal: they incur an excess risk $Ω\!\left(\frac{d-1}{n}\right)$, while there are simple non-rotation invariant algorithms with excess risk $O(\frac{s\log d}{n})$. The simplest rotation invariant algorithm is gradient descent on the logistic loss (with early stopping). A simple non-rotation-invariant algorithm for sparse targets that achieves the above upper bounds uses gradient descent on the weights $u_i,v_i$, where now the linear weight $w_i$ is reparameterized as $u_iv_i$.

Abstract: This study introduces a new normalization layer termed Batch Layer Normalization (BLN) to reduce the problem of internal covariate shift in deep neural network layers. As a combined version of batch and layer normalization, BLN adaptively puts appropriate weight on mini-batch and feature normalization based on the inverse size of mini-batches to normalize the input to a layer during the learning process. It also performs the exact computation with a minor change at inference times, using either mini-batch statistics or population statistics. The decision process to either use statistics of mini-batch or population gives BLN the ability to play a comprehensive role in the hyper-parameter optimization process of models. The key advantage of BLN is the support of the theoretical analysis of being independent of the input data, and its statistical configuration heavily depends on the task performed, the amount of training data, and the size of batches.

Volkswagen AG said Thursday that it will invest more than 2 billion euros ($1.94 billion) in a joint venture in China that aims to advance efficiency in automated driving. The German auto group said its Cariad software subsidiary will partner with Beijing-based Horizon Robotics, a supplier of smart-vehicle software solutions. Cariad will be the majority shareholder with a 60% stake, VW said, adding that the group will invest around EUR2.4 billion in the business. The JV will work to integrate numerous automated-driving functions in one chip, VW said. This will increase the stability of the system, saving costs and reducing energy consumption, VW said.

Around 5 billion won will be earmarked for another project to develop data and artificial intelligence-based products, the ministry said.

BERLIN, Nov 21 (Reuters) - German data mining software firm Celonis said on Thursday that it had raised $290 mln in a Series C funding round, putting a $2.5 billion valuation on the company that has been compared with enterprise application giant SAP . The funding round was led by Arena Holdings and investors included Ryan Smith, the founder of customer experience specialist Qualtrics that was bought by SAP for $8 billion a year ago. Celonis, based in Munich and New York, runs a cloud-based service that uses artificial intelligence to mine data and optimise business processes, serving customers including Siemens, 3M, Airbus and Vodafone. "We are in a market that shows enormous momentum," co-CEO and co-founder Bastian Nominacher told Reuters, adding that Celonis would invest the funds raised in its global sales and customer service and in enhancing its cloud platform. The funding round brings total investments into Celonis to $370 million.

German data mining software firm Celonis said on Thursday that it had raised $290 mln in a Series C funding round, putting a $2.5 billion valuation on the company that has been compared with enterprise application giant SAP . The funding round was led by Arena Holdings and investors included Ryan Smith, the founder of customer experience specialist Qualtrics that was bought by SAP for $8 billion a year ago. Celonis, based in Munich and New York, runs a cloud-based service that uses artificial intelligence to mine data and optimize business processes, serving customers including Siemens, 3M, Airbus and Vodafone. "We are in a market that shows enormous momentum," co-CEO and co-founder Bastian Nominacher told Reuters, adding that Celonis would invest the funds raised in its global sales and customer service and in enhancing its cloud platform. The funding round brings total investments into Celonis to $370 million.

Adopting'Big Data technologies' – including advances in analytics, machine-learning and artificial intelligence – could help the upstream sector save $75 billion per year by 2023, Wood Mackenzie said in a report on Monday. Digitalisation in the development of oil and gas fields could take the form of better processing of seismic data or new understanding of well logs and chemical analysis. This would help producers discover new resources in existing acreage, or give a competitive advantage when bidding in...

Omar bin Sultan Al Olama, the UAE's minister of state for Artificial Intelligence, and Deepak Bagla, the managing director and CEO of Invest India, signed a Memorandum of Understanding (MoU) to create a bilateral Artificial Intelligence Bridge that envisages the generation of $20 billion in economic benefits during the next decade. WAM said that the AI Bridge aims to spur discussion and explore options for the UAE and India to grow their Artificial Intelligence economies, according to an announcement after the signing ceremony by Invest India, the national investment promotion ad facilitation agency of the Indian government. The UAE Minister said that in the coming years, "how a country chooses to embrace Artificial Intelligence will have a tremendous impact on its ability to innovate and prosper. Data and processing will be a catalyst for innovation and business growth and serve as the backbone of more effective and efficient service delivery system." The designated Indian signatory said his country, which is "the world's fastest expanding market opportunity with its talent pool of human capital, well-acknowledged for innovation – and the UAE, a hub of cutting edge technologies – are natural partners in the field of Artificial Intelligence."