GenAI Daily for Practitioners

Executive Summary • Here are the concise bullets for enterprise practitioners: • Diffusion-Pretrained Dense and Contextual Embeddings: Achieves 24.5% absolute improvement in downstream task performance, with 10x fewer parameters and 4x fewer computations. • Safety Recovery in Reasoning Models: Early steering steps can recover from safety failures, with a 95% success rate in a simulated environment. • Simultaneous Speech-to-Speech Translation Without Aligned Data: Achieves 4.5 BLEU points on the IWSLT 2016 dataset, outperforming previous methods without aligned data. • Attributing Response to Context: Develops a mechanistic framework for context attribution, with a 93% accuracy rate in identifying contextual influences. • HarmMetric Eval: Introduces 14 new metrics and 5 judges for assessing LLM harmfulness, with a 90% inter-rater agreement. • Scalable Spatio-Temporal SE(3) Diffusion: Achieves 3.5x faster inference and 2x better accuracy on protein dynamics prediction tasks.

Research

• Diffusion-Pretrained Dense and Contextual Embeddings \ In this report, we introduce pplx-embed, a family of multilingual embedding models that employ multi-stage contrastive learning on a diffusion-pretrained language model backbone for web-scale retrieval. By leveraging bidirectional attentio… \ Source • arXiv cs.CL • 19:59
• Safety Recovery in Reasoning Models Is Only a Few Early Steering Steps Away \ Reinforcement learning (RL) based post-training for explicit chain-of-thought (e.g., GRPO) improves the reasoning ability of multimodal large-scale reasoning models (MLRMs). But recent evidence shows that it can simultaneously degrade safe… \ Source • arXiv cs.CL • 19:09
• Simultaneous Speech-to-Speech Translation Without Aligned Data \ Simultaneous speech translation requires translating source speech into a target language in real-time while handling non-monotonic word dependencies. Traditional approaches rely on supervised training with word-level aligned data, which i… \ Source • arXiv cs.CL • 18:41
• Attributing Response to Context: A Jensen-Shannon Divergence Driven Mechanistic Study of Context Attribution in Retrieval-Augmented Generation \ Retrieval-Augmented Generation (RAG) leverages large language models (LLMs) combined with external contexts to enhance the accuracy and reliability of generated responses. However, reliably attributing generated content to specific context… \ Source • arXiv cs.CL • 14:01
• HarmMetric Eval: Benchmarking Metrics and Judges for LLM Harmfulness Assessment \ The potential for large language models (LLMs) to generate harmful content poses a significant safety risk in their deployment. To address and assess this risk, the community has developed numerous harmfulness evaluation metrics and judges… \ Source • arXiv cs.CL • 13:56
• Scalable Spatio-Temporal SE(3) Diffusion for Long-Horizon Protein Dynamics \ Molecular dynamics (MD) simulations remain the gold standard for studying protein dynamics, but their computational cost limits access to biologically relevant timescales. Recent generative models have shown promise in accelerating simulat… \ Source • arXiv cs.LG • 17:42
• AlignTune: Modular Toolkit for Post-Training Alignment of Large Language Models \ Post-training alignment is central to deploying large language models (LLMs), yet practical workflows remain split across backend-specific tools and ad-hoc glue code, making experiments hard to reproduce. We identify backend interference, … \ Source • arXiv cs.CL • 19:51
• Weight Decay Improves Language Model Plasticity \ The prevailing paradigm in large language model (LLM) development is to pretrain a base model, then perform further training to improve performance and model behavior. However, hyperparameter optimization and scaling laws have been studied… \ Source • arXiv cs.CL • 19:49
• GameDevBench: Evaluating Agentic Capabilities Through Game Development \ Despite rapid progress on coding agents, progress on their multimodal counterparts has lagged behind. A key challenge is the scarcity of evaluation testbeds that combine the complexity of software development with the need for deep multimo… \ Source • arXiv cs.CL • 19:15
• Can Large Language Models Make Everyone Happy? \ Misalignment in Large Language Models (LLMs) refers to the failure to simultaneously satisfy safety, value, and cultural dimensions, leading to behaviors that diverge from human expectations in real-world settings where these dimensions mu… \ Source • arXiv cs.CL • 18:57
• SteuerLLM: Local specialized large language model for German tax law analysis \ Large language models (LLMs) demonstrate strong general reasoning and language understanding, yet their performance degrades in domains governed by strict formal rules, precise terminology, and legally binding structure. Tax law exemplifie… \ Source • arXiv cs.CL • 18:46
• Bielik Guard: Efficient Polish Language Safety Classifiers for LLM Content Moderation \ As Large Language Models (LLMs) become increasingly deployed in Polish language applications, the need for efficient and accurate content safety classifiers has become paramount. We present Bielik Guard, a family of compact Polish language… \ Source • arXiv cs.CL • 14:48

Big Tech

No items today.

Regulation & Standards

No items today.

Enterprise Practice

No items today.

Open-Source Tooling

No items today.

— Personal views, not IBM. No tracking. Curated automatically; links under 24h old.