World Wire

PDFs sit at the end of almost every workflow. Contracts, invoices, research papers, scanned intake forms, manuals, and compliance filings all arrive as PDFs. The format is very good at freezing a layout. It is much worse at letting you change one number inside that layout. That gap has become more visible as language models have gotten better at reading documents. A model can summarize a 90-page master services agreement in seconds. It will not reliably rewrite the source file. It will not hold the table alignment. It will not hand back a clean, signable copy. UPDF is built for the second half of that problem. It is an all-in-one PDF editor for Windows, macOS, iOS, and Android, developed by Superace Software Technologies . It combines editing, annotation, conversion, compression, OCR, page organization, and an integrated AI assistant in one application. Two Different Problems Living in the Same Document It helps to separate document work into two categories : The...

Liquid AI has released DSpark draft model checkpoints for three models in its LFM2.5 family: LFM2.5-1.2B-Instruct , LFM2.5-2.6B , and LFM2.5-8B-A1B . Each drafter adds a speculative decoding path to an existing target model. A roughly 300M-parameter draft proposes a block of nine candidate tokens, and the target model verifies the whole block in a single forward pass. The trade is a small memory increase for a large decoding speedup: up to 3.18x on an H100 and up to 2.87x on an M4 Max MacBook Pro. Output does not change. Under greedy decoding, the emitted sequence is identical to the target model running alone, so benchmark accuracy is unchanged. Both llama.cpp and SGLang have day-one support. Is it deployable? Yes, if you self-host. The weights ship as Safetensors and GGUF, and the drafter checkpoints are not served by any hosted inference provider on Hugging Face today. Running them needs an SGLang or llama.cpp build with DSpark support for LFM2 targets. Company l...

In this tutorial, we design an end-to-end preference-learning workflow using the Anthropic HH-RLHF dataset and Direct Preference Optimization (DPO). We begin by preparing a robust Colab environment, loading and parsing chosen–rejected response pairs, and auditing the dataset for structural and length-based preference biases. We then run lexical shortcut diagnostics to determine whether surface-level linguistic patterns can separate preferred from rejected responses, prepare conversational data with tokenizer-aware length filtering, and construct a version-robust DPO training pipeline with TRL and optional LoRA adaptation. Finally, we fine-tune a Qwen2.5-0.5B-Instruct model, evaluate reward accuracy and training behavior, analyze performance across individual HH-RLHF subsets, inspect potential length bias, generate sample responses, and save the resulting policy for further experimentation. Copy Code Copied Use a different Browser import dataclasses import importlib.util import in...