» Tag
nlp
26 postsEmpero AI Launches Qwythos-9B-v2: Looping Issues Resolved and Robustness Enhanced
Empero AI's Qwythos-9B-v2 resolves looping issues and enhances robustness, offering a 1M-token context for advanced applications.
Don't Let the LLM Speak, Just Probe Its Hidden State
LLMs already encode the answer in hidden states before generating text. A tiny probe turns this into fast, cheap, calibrated zero-shot classifiers without generation.
LFM2.5-Encoders: Fast Long-Context Inference on CPU
LFM2.5-Encoders released on Hugging Face offer fast performance for long-context inference on CPUs.
CommitBrief — AI code reviews, right in your terminal
A provider-agnostic, local-first CLI that reviews your staged changes, a historic range, or a whole GitHub pull request. Zero telemetry, no server. Free and open source.
commitbrief.comDKSplit – Fast Word Segmentation for Python
DKSplit is a Python library for fast word segmentation. Ideal for identifying brands and extracting keywords.
One encoder, seven heads: lessons from training a unified security classifier
We consolidated seven sequence classifiers into one multi-head model, sharing insights from our training process and results.
Show HN: Slopsift – A local, graph-backed linter for AI writing
SlopSift is a local graph-backed linter that identifies writing errors and filler content.
EU AI Act OpenRAG: 933 Legally Structured Chunks and BGE-M3 Embeddings
Explore the EU AI Act OpenRAG, featuring 933 structured chunks and BGE-M3 embeddings for legal-NLP experiments.
DeepSeek's new model sets a template for powerful LLMs that run lean
DeepSeek introduces Flash model 4.1, optimizing resource use while enhancing LLM capabilities.
1.5B Model Trained to Write Shell Commands on a Laptop CPU
User trains a 1.5B Qwen2.5-Coder model to generate shell commands. It operates on a laptop.
Claude Opus 5's Odd Behavior: Writing with Ambiguous Input
The odd behavior of Claude Opus 5 when faced with ambiguous inputs was discovered. Important implications for engineers.