korshunov.ai
English
All articles Page 1 / 90
media Hugging Face Forums · 9h ago

User Reports HuggingFace Charging for Unused L40S Compute in Spaces

A user on the Hugging Face discussion forum reported an issue where their Space remained stuck at the starting phase while using an L40S GPU. The user expressed frustration that they were being charged for compute resources despite the application failing to launch or utilize any actual processing power. This incident highlights concerns regarding billing transparency and infrastructure reliability within the platform's Spaces environment. The post serves as a complaint about financial loss due to technical failures rather than a feature announcement. No further technical details or official responses were included in the truncated source content.
media Hugging Face Forums · 9h ago

User Reports Step 3.7 Flash Model Tool Access Failure on HuggingChat

A user on the Hugging Face discussion forum reported that the Step 3.7 Flash model by StepFun AI has lost its ability to use tools, including MCP servers, as of the morning of the report. The individual expressed concern over whether this outage is temporary or permanent, noting their strong preference for this specific model due to its high performance and low resource costs compared to competitors. Despite praising the model's quality and affordability, the user highlighted the immediate disruption caused by the inability to execute tool-based functions. The post seeks clarification from the community regarding prior experiences with similar issues and potential resolutions. This incident underscores a critical dependency on tool availability for users relying on this specific AI configuration.
media Hugging Face Forums · 10h ago

Ontological Inversion: Flipping LLM Emotional Concepts via Negative Gain

The author introduces 'ontological inversion,' a technique designed to expand the one-directional inference nature of Large Language Models. This method allows models to capture nuanced, multifaceted concepts, such as memories that evoke both sorrow and joy simultaneously. The approach was developed by applying a negative gain factor during sweeps into the Niodoo steering architecture. It addresses the common limitation where LLMs overfit to singular emotional labels when prompted with personal experiences. By inverting concepts similarly to physics involution, the technique enables models to flip emotional states, such as transforming sorrowful memories into joyful ones. The work is shared via a GitHub repository titled 'ontological-inversion' by user Ruffian-L.
media Hugging Face Forums · 10h ago

User Inquires About Organization Rename Process on Hugging Face

A user posted on the Hugging Face discussion forum seeking assistance with renaming their organization. The individual stated they sent an email to website@huggingface.co on June 15 requesting a change from DZER-Studios to Vexion-LM. Despite sending the initial request, the user reported receiving no response and observed that the organization name remained unchanged. Consequently, the poster asked whether organization renames are still supported by the platform. They also requested guidance on alternative methods to contact the team regarding this specific administrative request.
media Hugging Face Forums · 10h ago

Community Inquiry on Model Benchmarking Methods

A user on the Hugging Face discussion forum posted a question seeking advice on how to benchmark machine learning models. The inquiry was initiated by an individual who is new to the field of fine-tuning and wishes to evaluate their models after completion. The post explicitly asks for established methods or strategies that the community uses for this purpose. It highlights a common need among practitioners to understand standard evaluation practices in model development. The discussion thread currently contains only one post from a single participant. No specific benchmarks, metrics, or technical solutions were provided within the visible content of the source.
media Hugging Face Forums · 10h ago

Qwen3/Gemma3 Candle Skips Attention Masks for Equal-Length Batches in CPU Mode

A user has reported a critical bug in the Hugging Face text-embeddings-inference library affecting Qwen3 and Gemma3 models. The issue arises when running inference on CPUs with concurrent requests, leading to significant accuracy degradation. Specifically, the Candle backend incorrectly skips attention masks for batches where all input sequences have equal lengths. This defect compromises the reliability of embeddings generated under these specific conditions. To address the problem, the author submitted a pull request containing a fix that was thoroughly tested on their local machines. The bug highlights potential stability risks in CPU-based embedding services handling batched inputs.
github LlamaIndex · 10h ago

Llama Index v0.14.23 Release Notes

Llama Index released version 0.14.23 on June 24, 2026, introducing significant multimodal capabilities and various bug fixes. The core update includes multimodal synthesis features and the introduction of multimodal query engines to support diverse data types. Key fixes address document and video block handling within FunctionTool outputs and ensure URL-backed memory blocks are preserved correctly. Performance improvements were implemented by using sets for within-batch deduplication in the ingestion pipeline and optimizing token text splitting logic. The release also resolves a ZeroDivisionError on empty input sequences and fixes recursion errors in splitters when units exceed chunk sizes. Additionally, explicit UTF-8 encoding was added to file I/O operations, and deep copying of initial states prevents mutation leaks across workflow runs.
lab Claude Code Releases · 10h ago

Claude Code v2.1.191 Release Notes

Claude Code version 2.1.191 introduces /rewind support, allowing users to resume conversations from before a /clear command was executed. The update fixes several critical issues, including background agents resurrecting after being stopped and scroll position jumping during streaming responses. It also corrects behavior where /voice displayed generic error messages and where /login URLs were truncated in Windows Terminal. Significant improvements enhance reliability for MCP servers by adding retry logic for transient network errors during capability discovery and OAuth flows. Headless environments now skip browser popups for OAuth, while sandbox network permissions are remembered for the session duration. Performance optimizations reduce CPU usage during streaming by approximately 37% through text update coalescing and mitigate long-session memory growth from the terminal output cache.
arxiv arXiv cs.LG · 12h ago

AdaR: Adaptive Recurrent Message Passing for Graph Test-Time Computing

AdaR enables flexible test-time computing on graphs without parameter changes by using adaptive recurrence. It derives step dependence as a necessary and sufficient condition for convergence and incorporates normalized step information and representation-target relations into recurrent updates, guided by gradient-based supervision signals. Empirical results show AdaR outperforms strong baselines in both inductive and transductive graph learning settings.
arxiv arXiv cs.LG · 12h ago

Speech-Text Models Latently Transcribe Speech in Intermediate Layers

Interleaved speech-language models undergo an implicit transcription phase where spoken words become decodable as text tokens in intermediate layers, despite no speech recognition training. Up to 77% of the data shows the spoken word appearing as a top candidate text prediction, followed by a transition to text-based next-word prediction before returning to speech. This behavior is influenced by interleaved training and text LM initialization, and correlates with spoken knowledge performance.
arxiv arXiv cs.LG · 12h ago

LLM-Integrated App Bug Seams Reveal Testing Gaps

A rental-search assistant with LLM features and multi-market support faced persistent user defects despite 1,553 passing automated tests. Analysis of 252 bug-fix commits showed 44% of fixes occurred at four unseen seams: browser runtime, non-default market, end-to-end flows, and whole-system level. A fix without a seam guard caused a defect to ship twice, highlighting the need for targeted testing at these boundaries.
arxiv arXiv cs.LG · 13h ago

Deep Learning Pipeline for Sign Language Recognition and Translation to Indian Vernaculars

A two-stage deep learning model classifies Indian sign language video clips into English words using a fine-tuned VideoMAE transformer, achieving 99% training and 78% validation accuracy on a 13-class dataset. The predicted English labels are translated into Hindi, Telugu, and Bengali using Meta AI's NLLB-200 multilingual model, with a Streamlit demo enabling user-uploaded video inference and cross-lingual output.
arxiv arXiv cs.LG · 13h ago

The Scissors Effect: Resize Diversity Hurts Robust Surrogate Transfer

Input diversity, a common practice in transfer attacks, improves success on standard surrogates but reduces it on robust ones. This regime-dependent effect, called the Scissors Effect, is driven by gradient geometry, with resize operations degrading alignment in robust models. A training-free rule (CG-DI) adjusts diversity based on local gradient consistency to preserve attack success across surrogate types.
arxiv arXiv cs.LG · 13h ago

HERTA: Automated Testing for FHE Framework Vulnerabilities

HERTA is the first automated testing tool designed for fully homomorphic encryption frameworks. It uses metamorphic testing with novel relations derived from FHE semantics to detect deep-seated logic bugs that can silently corrupt encrypted computations. Evaluation on three industry frameworks revealed 21 previously unknown bugs, several of which have been confirmed and fixed by developers, with significant implications for security and service integrity.