Search: “benchmark”

Qwen3.8-27B uncensored Q6_K at 156K context on one RTX 5090, 140-190 tok/s with DFlash2

Setup for one long agentic coding session (tools + vision) on a single 5090: Qwen3.8-27B RVN Heretic (ARA abliterated) at Q6_K, 159744 context, 1 slot, q8_0 K/V, DFlash2 speculative decoding, vision p

Content type

All Blog X Post Paper YouTube Reddit GitHub

Clear filters

Model Releases

Racing in Volume with Flow Ensembles

DGX agent

arXiv:2609.16310v1 Announce Type: new Abstract: Streaming 4D reconstruction has been demonstrated only indoors, on dense camera rigs surrounding subjects that move at human pace. Outdoor 4D reconstruc

model-releasesarxiv-cs-cv

16 Sept 2026

Local Ai

[Release] SOTA GGUFs for Qwen3.8-Flash-Next: GSQ-RCO Providing Near Baseline Performance

https://preview.redd.it/e5wn8eyh7vph1.png?width=1080&format=png&auto=webp&s=5faeec866acc9e8eca8dc9e84a6661bee7bcf954 https://preview.redd.it/8ov5gl8j7vph1.png?width=1080&format=png&auto=webp&s=440320c

local-air-localllama

Local Ai

RepoAtlas: Guiding Coding Agents via Evolving Multimodal Repository Views

arXiv:2609.16936v1 Announce Type: cross Abstract: Large language model (LLM)-powered coding agents have made rapid progress in automating software engineering tasks, yet repository-level issue resolut

local-aiarxiv-cs-ai

ResLRP: The Role of Residual Cancellation in Attribution Instability in Vision Transformers

arXiv:2609.17152v1 Announce Type: cross Abstract: Vision Transformers (ViTs) are central to most modern vision models, yet obtaining input attributions that are fine-grained, faithful, and stable rema

Right Tool, Right Job: Native-Language Evaluation, Tokenizer Sensitivity, and Methodological Findings from a French-Only BabyLM

arXiv:2609.17435v1 Announce Type: new Abstract: We submit MeTRON-FR, a 125M GPT-2 pretrained on 92.47M words of French, to the BabyLM 2026 Strict track. It scores 85.97 +/- 0.17% on QFrBLiMP (a native

Safe Error Correction for Language Models: Frozen-Base Adjustment with Capability Preservation

arXiv:2609.16145v1 Announce Type: new Abstract: We study a practical question: can a small correction module fix errors in a frozen language model's outputs without degrading its base capabilities? We

Same Problem, Different Field: Cross-Domain Solution Import via Domain-Stripped Computational Fingerprints

arXiv:2609.07595v2 Announce Type: replace-cross Abstract: The same underlying computational problem is solved across unrelated fields under different names: recursive Bayesian state estimation appears

Style-Debiased DPO: Updating LLM Knowledge with Factuality-Aware Synthetic Preference Data

arXiv:2609.16532v1 Announce Type: new Abstract: Continued pretraining (CPT) with data augmentation such as paraphrasing can store inside a large language model (LLM) the knowledge of a small source co

the design of http://evaluatorbench.com grew out of a summer project, http://epistemedia.org (which i still need to write about), and runs o…

model-releasesyohei-nakajima--x

the design of http://evaluatorbench.com grew out of a summer project, http://epistemedia.org (which i still need to write about), and runs on @activegraphai the pattern across all three projects is th

Safety

The essay below from Aaronson is just unmoored from reality, sorry. It is just not true, if you try to approach the question in a data-orien…

The essay below from Aaronson is just unmoored from reality, sorry. It is just not true, if you try to approach the question in a data-oriented fashion, rather than via vibes and speculative fiction a

safetygary-marcus--x

Universal Defenses for Tool-Integrated LLM Agents Against Adversarial Attacks

arXiv:2609.16098v1 Announce Type: cross Abstract: Large Language Model (LLM) agents have demonstrated impressive capabilities across a variety of domains, particularly when integrated with external to

Upgraded my local setup with 2 rtx pros and it's amazing.

Follow up post of https://www.reddit.com/r/LocalLLaMA/s/nGMyKswrch. Thanks everyone who replied. I didn't change the specs. Might be loosing some of the memory bandwidth but will scale in future if I

Where Should a Document Live: Context, Representations, or Parameters?

arXiv:2609.17346v1 Announce Type: cross Abstract: To answer questions outside of their pre-training data, large language models (LLMs) need access to new information, which can be presented in the con

Write up on building and improving reliability of agentic harnesses + a big dataset from my work.

I've been working on something for this community (and other self-hosted communities) for almost 5 months now. Personally, I would grade it as probably the closest open source equivalent of Manus and

XeBoostLM: Native C++ Local LLMs for Intel NPUs & iGPUs

https://reddit.com/link/1wi2zuc/video/rs1a4sx92xph1/player Hey everyone, I’ve been experimenting with local LLMs on Intel hardware, and I wanted a hyper-optimized, standalone CLI tool that didn't rely

A parameter-efficient three-branch architecture for multimodal misinformation detection with limited annotations

arXiv:2512.20257v2 Announce Type: replace Abstract: With the rise of easily accessible generative tools for creating and manipulating multimedia content, the threat of realistic synthetic alterations

Agents

A primer on evaluation methods for large language models in healthcare

arXiv:2609.14819v1 Announce Type: cross Abstract: Large language models (LLMs) have a growing range of applications in medicine, and their evaluation is critical for ensuring they provide benefit and

agentsarxiv-cs-ai

Adaptive Conformal Redistribution for Inter-class Transitional Uncertainty in Medical Image Classification

arXiv:2609.13303v1 Announce Type: cross Abstract: Medical image classification is frequently complicated by transitional categories whose feature distributions overlap those of adjacent classes, produ

Research

All Learning Has an Emotional Basis, So Does Task-Oriented Dialogue

arXiv:2507.01594v2 Announce Type: replace Abstract: Task-oriented dialogue (ToD) systems aim to help users accomplish goals through natural language interaction. Beyond task success, effective ToD sys

researcharxiv-cs-cl

Authorship attribution and aesthetic evaluation of AI poetry: a case study with Haiku

arXiv:2609.15511v1 Announce Type: cross Abstract: This paper investigates the generation and human evaluation of Japanese haiku by contemporary Large Language Models (LLMs), focusing on authorship per

Applications

Beyond OCR Accuracy: Text-Centric VQA Under Image Degradation with Modular and End-to-End

arXiv:2609.13815v1 Announce Type: new Abstract: Text-centric Visual Question Answering (VQA) requires reading and reasoning over text embedded in images, a task made substantially harder when images s

applicationsarxiv-cs-cv

Agents

DARE: Dialectical Agentic Reasoning for Structured Knowledge Fact Checking

arXiv:2609.13808v1 Announce Type: new Abstract: Structured knowledge fact checking aims to determine the truthfulness of natural language claims by reasoning over structured evidence. Recent program-g

agentsarxiv-cs-cl

DeepSeek V4.1F Q4 on M3 Ultra with native DSpark MTP (40tps / 800tps)

I liked DeepSeek V4.1 Flash as an agent model but at 16 t/s on ds4 it was painful to sit through a real turn. It seemed some redditors and m3 ultra owners appreciated my glm 53 flash optimizations, so

DiffuTester: Accelerating Unit Test Generation for Diffusion LLMs via Mining Structural Pattern

arXiv:2509.24975v4 Announce Type: replace-cross Abstract: Diffusion large language models (dLLMs) enable parallel generation and are promising for unit test generation (UTG), where efficient and large

Do Tabular Foundation Models Still Need Feature Engineering?

model-releasesarxiv-cs-lg

arXiv:2609.13202v1 Announce Type: new Abstract: Feature engineering has long been a cornerstone of tabular machine learning. Tabular foundation models (TFMs) are pretrained on a wide range of tabular

Does Attention-Guided Masking Really Help Object Discovery in Object-Centric Learning?

arXiv:2609.15187v1 Announce Type: new Abstract: Object-Centric Learning (OCL) aims to decompose images into objects without human annotations. A major family of mainstream methods uses Slot Attention

Empirical Evaluation of Open-Source Large Language Models for Retrieval-Augmented Generation in ESG Domain

arXiv:2609.15242v1 Announce Type: new Abstract: Environmental, Social, and Governance (ESG) reporting is critical for corporate accountability, with Large Language Models (LLMs) and Retrieval-Augmente

Evaluating Context Segmentation in Locally Deployable SLMs for Cybersecurity CTF Tasks

arXiv:2609.12839v2 Announce Type: replace-cross Abstract: The proliferation of highly capable open-weight Small Language Models (SLMs) democratizes access to advanced cybersecurity capabilities, posin

GorillaWatch: An Automated System for In-the-Wild Gorilla Re-Identification and Population Monitoring

arXiv:2512.07776v2 Announce Type: replace Abstract: Monitoring critically endangered western lowland gorillas is currently hampered by the immense manual effort required to re-identify individuals fro

GRAVA: Grounded Reasoning-to-Action Representation and Learning for Autonomous Driving

arXiv:2609.15169v1 Announce Type: new Abstract: Driving vision-language-action (VLA) models increasingly reason before acting, but their intermediate reasoning is often weakly grounded in physical sce

Safety

HISPO: Hierarchical Importance-Sampling Policy Optimization with Entropy-Derived Segments

arXiv:2609.15471v1 Announce Type: new Abstract: Reinforcement learning with verifiable rewards (RLVR) has become a central approach for improving mathematical reasoning in language models, but long-fo

safetyarxiv-cs-ai

Safety

Hypergraph-Regularized Gramian Volumes for Multimodal Retrieval

arXiv:2609.15320v1 Announce Type: new Abstract: Volume-based multimodal retrieval jointly scores a text query with a candidate's video, audio, and subtitle embeddings. While this approach captures hig

safetyarxiv-cs-cv

If you have a 3090, or other 30xx for local LLMs, I have something for you

I have a custom fork of llama.cpp designed around the ampere architecture specifically (though many of the upgrades also translate to faster performance of blackwell + lovelace). The recommended confi

Issue Bias in Generative AI Writing Assistance: Political Issues and LLMs in the Swedish 2026 Election

arXiv:2609.15207v1 Announce Type: new Abstract: Generative AI writing assistants and the Large Language Models (LLMs) that power them are increasingly part of how voters gather information before elec

JaxAHT: A JAX-Based Library for Ad Hoc Teamwork

arXiv:2609.13716v1 Announce Type: new Abstract: Ad Hoc Teamwork (AHT) addresses the challenge of designing agents capable of coordinating with novel partners without prior coordination. However, progr

LPA-CWM: A Learned Physical Adjudicator for Motion Reasoning with Counterfactual World Models

arXiv:2609.14073v1 Announce Type: cross Abstract: Counterfactual world models (CWM) extract motion from pretrained video predictors by comparing factual and intervened predictions. However, responses

MemRiskBench: Trace-Aware Risk-Preserving Evaluation for Long-Horizon LLM Agents

arXiv:2609.14976v1 Announce Type: new Abstract: Long-horizon LLM agents accumulate memory across sessions, creating sparse but high-impact risks: stale facts, conflicting updates, cross-user leakage,

Modality-Guided Mixture of Structured Experts with Entropy-Triggered Routing for Multimodal Recommendation

arXiv:2602.20723v3 Announce Type: replace Abstract: Multimodal recommenders combine collaborative behavior with visual and textual item evidence, whose usefulness varies across user-item interactions.

MoME: Mixture-of-Memory Embeddings for Context-Aware Sparse Lookup

arXiv:2609.15126v1 Announce Type: cross Abstract: Scaling large language models efficiently has motivated sparse capacity mechanisms such as Mixture-of-Experts and, more recently, conditional memory:

N^2: A Unified Python Package and Test Bench for Nearest Neighbor-Based Matrix Completion

model-releasesarxiv-cs-lg

arXiv:2506.04166v3 Announce Type: replace Abstract: Nearest neighbor (NN) methods have re-emerged as competitive tools for matrix completion, offering strong empirical performance and recent theoretic

Neyshekar: An Open Persian Read-Speech Corpus for Automatic Speech Recognition

arXiv:2609.14542v1 Announce Type: new Abstract: Neyshekar is presented as an open Persian read-speech corpus designed for coverage of both formal and informal language, named entities, and longer utte

Agents

North Small Translate: Advanced Cost-Effective Translation (Cohere CAT+)

arXiv:2609.13916v1 Announce Type: new Abstract: We present North Small Translate, an open-weight, LLM-based machine translation (MT) model with instruction-following capabilities built on the same fou

agentsarxiv-cs-cl

Occlusal Geometry in Closed Form for Orthodontic Report Generation

arXiv:2609.13237v1 Announce Type: cross Abstract: Orthodontic report generation from intraoral data is normally cast as multimodal captioning, yet the released Bite2Text scan pairs are supplied alread

PACE: Progressive Angular-to-Norm Contrastive Embedding

arXiv:2609.15152v1 Announce Type: cross Abstract: Multimodal embedding models encode heterogeneous inputs into a shared embedding space, enabling efficient similarity computation across modalities and

Parameter-Efficient Quantum NLP for Paraphrase Detection: Performance, Robustness, and Entanglement

arXiv:2609.14529v1 Announce Type: cross Abstract: Rigorous empirical validation of quantum machine learning on natural language tasks remains scarce. We evaluate a 10-qubit hybrid quantum-classical va

PhysBrain 1.5: From Vision-Language Models to Physical Foundation Models

arXiv:2609.14973v1 Announce Type: new Abstract: We present PhysBrain 1.5, a unified model for understanding physical environments, generating actions, and predicting future states. Motivated by the ph

PIVOT: Physics-Grounded Verification for AI-Generated Audio-Video Detection

arXiv:2609.15562v1 Announce Type: cross Abstract: As generative models continue to advance, AI-generated content (AIGC) is becoming increasingly realistic, weakening the artifact cues commonly exploit