←Search results

Search “GitHub”

By relevance · Total 1629 results Clear search

Saturday

  3. PyTorch Blog

    PyTorch introduces Zephon elastic deterministic data loading for TorchTitan

    The PyTorch blog introduces Zephon, bringing elastic deterministic data loading to TorchTitan: after training a 1B-parameter model to step 150 on 16 GPUs, training was resumed on 8, 32, and 64 GPUs, and Zephon's loss curves overlapped exactly, with the SHA-256 of every global batch matching the baseline. With the standard data loader, training the same 1B model on 20B tokens across 8 to 64 GPUs produced fluctuations of up to 0.82 points on DCLM Core v1 and 0.64 points on FineWeb, whereas Zephon showed only 0.05 and 0.011 points. Zephon performs tokenization, packing, shuffling, and mixing on the fly within the training job, with no need for preprocessed corpus copies.

    The PyTorch blog introduces Zephon, bringing elastic deterministic data loading to TorchTitan: after training a 1B-parameter model to step 150 on 16 GPUs, training was resumed on 8, 32, and 64 GPUs, and Zephon's loss curves overlapped exactly, with the SHA-256 of every global batch matching the baseline. With the standard data loader, training the same 1B model on 20B tokens across 8 to 64 GPUs produced fluctuations of up to 0.82 points on DCLM Core v1 and 0.64 points on FineWeb, whereas Zephon

    Matched fieldsBody

    #engineering#English text is being prepared.#English text is being prepared.#English text is being prepared.

Friday

  2. Databricks Blog

    Databricks uses domain knowledge injection to improve LLMs' ability to detect malicious npm binary packages

    Since November 2025, Databricks has been collecting malicious npm samples. Among the 11,413 package-versions scanned and verified, 96.2% hid their payloads in source code or on external servers, and only 3.8% (432) bundled binaries; however, the proportion of such packages for which no clear malicious conclusion could be reached was more than twice that of pure source-code packages (14.4% vs 6.2%). The team evaluated 9 models (including Claude Haiku 4.5, Gemini 3.5 Flash, GPT-5.4-nano, Qwen3-Next 80B, etc.) and injected precomputed static evidence as capability profiles during inference to improve detection quality for both budget-tier and frontier models.

    image1.png

    Since November 2025, Databricks has been collecting malicious npm samples. Among the 11,413 package-versions scanned and verified, 96.2% hid their payloads in source code or on external servers, and only 3.8% (432) bundled binaries; however, the proportion of such packages for which no clear malicious conclusion could be reached was more than twice that of pure source-code packages (14.4% vs 6.2%). The team evaluated 9 models (including Claude Haiku 4.5, Gemini 3.5 Flash, GPT-5.4-nano, Qwen3-Nex

    Matched fieldsBody

    #company#Safety / alignment#Agents#English text is being prepared.
  4. Unite.AISelected

    Reflection open-sources 501B-parameter MoE model Beam under Apache 2.0

    On October 16, Reflection released its first open-weight model Beam on Hugging Face under the Apache 2.0 license, and Baseten provided day-0 support the same day through Model APIs and Dedicated Inference. Beam is a 501B total-parameter MoE autoregressive Transformer with 23B activated per token, a context window of up to 1 million tokens, pretraining on 23.8 trillion tokens, and weights released in BF16, FP8, and NVFP4 precisions. The evaluation results given by Reflection are GPQA Diamond 90.5%, SWE-bench Verified 80.9%, AIME 2026 97.8%, Terminal-Bench 2.1 80.1%, and MCP Atlas 78.7%, and it also open-sourced the terminal agent tool Mirror CLI.

    On October 16, Reflection released its first open-weight model Beam on Hugging Face under the Apache 2.0 license, and Baseten provided day-0 support the same day through Model APIs and Dedicated Inference. Beam is a 501B total-parameter MoE autoregressive Transformer with 23B activated per token, a context window of up to 1 million tokens, pretraining on 23.8 trillion tokens, and weights released in BF16, FP8, and NVFP4 precisions. The evaluation results given by Reflection are GPQA Diamond 90.5

    Matched fieldsBody

    #hot-signal#Model releases#Open-source ecosystem#Agents
  6. Claude Blog

    How an Anthropic Japan BDR Uses a Team of Claude Agents to Handle Daily Work

    An Anthropic Japan business development representative used Claude Managed Agents (beta) to build an agent team called Nakama, handing lead triage, qualification, email drafting, Salesforce updates, and post-meeting follow-up to 6 agents that run on a schedule, completing most of the work before the start of each day. He first prototyped it in Claude using skills, scheduled tasks, and a shared folder; after migrating, skills became system prompts, the shared folder became memory storage, and changes are proposed uniformly in a Slack channel and synced by a coordinating agent roughly every 15 minutes.

    An Anthropic Japan business development representative used Claude Managed Agents (beta) to build an agent team called Nakama, handing lead triage, qualification, email drafting, Salesforce updates, and post-meeting follow-up to 6 agents that run on a schedule, completing most of the work before the start of each day. He first prototyped it in Claude using skills, scheduled tasks, and a shared folder; after migrating, skills became system prompts, the shared folder became memory storage, and cha

    Matched fieldsBody

    #lab#Tutorials / practice#Agents#English text is being prepared.
  11. 36Kr · AISelected

    DeepMind AlphaProof Nexus appears in《Science》: autonomously solves 9 Erdős problems

    Google DeepMind's automated mathematical proof system AlphaProof Nexus appeared in the October 8 issue of《Science》, autonomously solving 9 of 353 open Erdős problems, two of which had remained open since 1970. The system uses Gemini 3.1 Pro to draft proofs in Lean, with a compiler verifying line by line and feeding back iterations; the reasoning cost per problem is several hundred dollars. The paper's ablation results show that a simplified system that removes complex structures such as the sub-agent tournament and retains only the large model plus Lean feedback can also solve these 9 problems, though the hardest few are more expensive.

    Google DeepMind's automated mathematical proof system AlphaProof Nexus appeared in the October 8 issue of《Science》, autonomously solving 9 of 353 open Erdős problems, two of which had remained open since 1970. The system uses Gemini 3.1 Pro to draft proofs in Lean, with a compiler verifying line by line and feeding back iterations; the reasoning cost per problem is several hundred dollars. The paper's ablation results show that a simplified system that removes complex structures such as the sub-

    Matched fieldsBody

    #hot-signal#Papers/Research#Reasoning#DeepMind

Thursday