Modern Standard Arabic
PulseAugur coverage of Modern Standard Arabic — every cluster mentioning Modern Standard Arabic across labs, papers, and developer communities, ranked by signal.
3 day(s) with sentiment data
-
New ASAS benchmark reveals significant safety gaps in Arabic LLMs
A new benchmark called ASAS has been developed to evaluate the safety of Arabic large language models (LLMs). The benchmark, which includes 801 human-curated prompts across eight safety categories, revealed that most te…
-
New Arabic Dialect Safety Benchmark Released
Researchers have introduced ArabicDialectSafety, a new dataset designed to evaluate the safety of Arabic content across various dialects. The dataset includes 25,071 prompts in six Arabic varieties and is annotated for …
-
MiniMax AI and Fireworks AI Open-Source Inference Kernels
MiniMax AI and Fireworks AI have jointly released open-source kernel work aimed at accelerating inference speeds. This collaboration focuses on making the underlying computational components accessible, emphasizing that…
-
Minimax M3 vision and Arabic support merged into llama.cpp
The Minimax M3 large language model now has vision support integrated into the llama.cpp library. This merge, initiated by a pull request, also includes support for Modern Standard Arabic (MSA) within the llama.cpp fram…
-
SemiAnalysis: Jensen's committee approach hinders US open-source AI progress
SemiAnalysis criticizes Jensen Huang's committee-based training approach for open-source AI, arguing it stifles innovation and leads to groupthink. The analysis suggests that independent competition among labs, similar …
-
New CTC method improves diacritic restoration for Arabic speech transcripts
Researchers have developed an efficient, non-autoregressive method for diacritic restoration in Arabic speech transcripts using Connectionist Temporal Classification (CTC). This approach incorporates hard constraints du…
-
Arabic LLMs can be steered to dialects using neuron analysis
Researchers have explored methods to steer Arabic Large Language Models (LLMs) towards generating specific dialects without requiring fine-tuning. The study identified sparse neuron populations that encode dialect-speci…
-
Arabic LLMs can be steered to use specific dialects without retraining
Researchers have developed methods to control the dialectal output of Arabic language models without requiring additional training. By analyzing neuron-level representations, they identified specific neuron populations …
-
OCI's AI scaling challenges and cost-reduction tools emerge
Oracle Cloud Infrastructure (OCI) is facing challenges in scaling AI workloads, as highlighted by the limitations of its Modern Standard Arabic (MSA) offering. While AI agents are contributing to database sprawl, they a…
-
New research explores hybrid and sparse attention mechanisms for LLMs
Researchers are exploring novel methods to optimize attention mechanisms in large language models, particularly for handling long contexts. The HydraHead architecture, for instance, hybridizes Full Attention (FA) and Li…
-
MiniMax AI Celebrates Community Contributions to M3 Open Weights
MiniMax AI is celebrating the open-source community's work with its M3 open weights. The mlx-vlm project has integrated support for MiniMax M3, including an implementation for Modern Standard Arabic. This integration wa…
-
MiniMax AI releases open-source M3 model with 1M context
MiniMax AI has released its new open-source model, MiniMax M3, featuring advanced coding and agentic capabilities. The model supports native image and video input, can interact with computers, and boasts a 1 million tok…
-
MiniMax AI releases open-weight M3 model with 1M context
MiniMax AI has released its new open-weight model, MiniMax M3, featuring a 1 million token context window and advanced capabilities. The model utilizes a novel sparse attention architecture called MSA, which includes de…
-
Fireworks AI launches MiniMax M3 inference with multimodal features
Fireworks AI has launched an inference infrastructure service for the MiniMax M3 model, offering Day-0 access and competitive pricing. This new service boasts multimodal capabilities, including slide processing and gene…
-
MiniMax AI open-sources MSA kernel library, M3 weights coming Friday
MiniMax AI has announced the open-sourcing of its high-performance MSA kernel library. The company also stated that the M3 weights are scheduled for release this Friday. This release includes links to the library's GitH…
-
MiniMax M3 launches with 1M-token context and MSA architecture
MiniMax has released its M3 model, featuring a novel Sparse Attention (MSA) architecture that enables a 1 million token context window and native multimodality. This new architecture significantly reduces computational …
-
New dataset benchmarks LLMs on cultural reasoning in Arabic dialogues
Researchers have developed a new dataset, ArabCulture-Dialogue, to address the lack of culturally rich conversational data for evaluating Large Language Models (LLMs) in Arabic. This dataset covers 13 Arabic-speaking co…
-
LLMs can now generate Arabic poetry with specific style and rhyme controls
Researchers have developed a new method for instruction-guided poetry generation in Arabic and its dialects. They created a large dataset of instruction-based poetry examples in Modern Standard Arabic and various dialec…
-
New MSA algorithm uses model history to efficiently unlearn data from LLMs
Researchers have introduced a novel algorithm called Model State Arithmetic (MSA) designed to efficiently remove the influence of specific data points from large language models. Unlike complete retraining, which is com…