Category
3,113 papers
Continued pretraining (CPT) with data augmentation such as paraphrasing can store inside a large language model (LLM) the knowledge of a small source corpus. The stored knowledge, however, is not always retrieved correctly. We study the eliciting side rather than the storing side: we use preference optimization, which...
De-identified r\'esum\'e screening assumes that redacting explicit fields prevents ethnocultural inference; however, recent audits attribute residual leakage to declared languages. We investigate whether eliminating language fields resolves this leakage across nine open-weight models and 620 counterfactual r\'esum\'es....
Masked diffusion language models (dLLMs) generate text by iteratively denoising masked tokens with bidirectional attention. Extending reasoning across generation chunks normally requires keeping earlier generated text in context. We ask whether a dLLM can instead continue reasoning after that text is cleared, using onl...
We study a practical question: can a small correction module fix errors in a frozen language model's outputs without degrading its base capabilities? We propose CRN v2, a lightweight logit-level correction module (~34M trainable parameters, 0.73% of the 4.65B text module) that sits atop a fully frozen Gemma 4 E2B model...
Reach audiences
Advertise in front of researchers, engineers, and readers.
Coupled Calibration and Learning (CCL) is proposed, an LLM distillation algorithm that couples teacher calibration with student updates through token-level branching, using reward feedback only on source questions.
We present an efficient method to distill reasoning capabilities into compact video-language models (VLMs) for video question answering (VideoQA). Our approach fine-tunes a 2B-parameter model using only $\sim$900 uncertainty-selected examples, each augmented with synthetic chain-of-thought (CoT) rationales generated by...
ActProbe, an internal-state-based framework for detecting and localizing poisoned segments in multi-source LLM inputs, is proposed and remains effective against defense-aware adaptive attacks and can protect black-box APIs through surrogate-based poisoned-segment removal.
The proposed framework uses a shared-specific Mixture-of-LoRA architecture comprising one shared LoRA and six task-specific expert LoRAs, together with a two-stage training procedure, and achieves good results in the FLARE 2026 Task 3 test sets.
Testing on a large cohort of soil metagenomes, it is found that BBERT identifies bacterial sequence syntax without relying on reference databases, enabling accurate assignment of taxonomic domain, coding potential, and reading frame directly from reads as short as 100 bp.
This is unedited work with Mistral Vibe. Quality/logic checked through multiple different instance reviews. Proceed with caution. This manual presents a novel methodology for detecting algorithmic bias in large language models (LLMs) by analyzing GPU power-line current fluctuations during inference. Unlike existing bia...
Ternary (1.58-bit) language models are stored in practice at two bits per weight, 26% above the information content of a ternary symbol. The code that removes that redundancy is not new: five ternary values as the base-3 digits of one byte at 1.600 bits per weight is llama.cpp's TQ1_0 and Spectra 1.1's TQ1; the depth-o...
Pre-Alpha Ethiopic (Geʽez / Amharic) recognition model — PP-OCRv6 medium fine-tune Status: preliminary / pre-Alpha. This deposit is an early public checkpoint, not a finished production model. Better versions are planned as soon as additional medieval and pre-modern Ethiopic ground truth can be annotated and curated. H...
A new computational approach reveals how cells communicate to assemble into tissues. This could accelerate stem cell engineering, improve disease models, and inform therapies.
Zero-knowledge proofs can verify computing results without rerunning workloads or exposing sensitive data. Magnetar helps bring them into the mainstream with support for standard programming languages, major speed gains, and quantum-resistant security. The post Proof-based general-purpose computing with Magnetar appeared first on Microsoft Research.
Learn how Google DeepMind and GovTech Singapore uses lightweight on-device Gemma models to intercept conversational scams in private chats with zero cloud data egress.
StudentSim is a research system for evaluating and improving AI tutoring tools before they are tested with learners. It models how individuals may engage with tutors, giving feedback to improve tutor performance across subjects like chess, writing, and mathematics. The post Simulating student learning for better AI tutoring appeared first on Microsoft Research.
We use cookies to run the site and, with your consent, for analytics and to show ads. See our Cookie Policy.
