AI Wiki — Knowledge Base

Tuned/abliterated Qwen3.8-27b into a 24gb card 262k guff using the newest unreleased version of LexiPanel. It's fast with reliable draft acceptance. Made for 7900xtx but should work on whatever 24gb card with this setup and headless. Doesn't get dumber while coding like most of the other fine-tunes.

https://huggingface.co/Wa1k3r/Qwen3.8-27b-CODER-4q_xs-24GB-262k-Optimalcardfit Qwen3.8-27B CODER — IQ4_XS imatrix · 24 GB card fit · ~262k context · MTP draft Quantized, Abliterated, and fitted by Lex

model-releasesr-localllama

2 Oct 2026

Search: “reasoning”

R-GroundBench: A Diagnostic Benchmark for R-Group Groundingin Markush Molecular Editing

Spatial Lifting for Dense Prediction

Beyond Layers: Position-Resolved Gradient Conflict and Position-Aware Modulation for Unified Multimodal Models

Beyond Prediction: Steering VLM Agents with Retrospective World Modeling

Cognitive Enhancement: Rethinking the Necessity of Role-Playing for Large Language Models

CombEval: A Framework for Evaluating Combinatorial Counting in Large Language Models

Composition, Not Conversation: VLMs Lose the Scene, Not the Thread

Consistent Plan-Act for Long-Horizon Agentic Tasks

E2E-SWE: Benchmarking LLMs on Building Working Codebases from Scratch

EvoGuard: An Extensible Agentic RL-based Framework for Practical and Evolving AI-Generated Image Detection

Front-to-Back: Benchmarking Vision-Language Models for Asymmetric Cross-View Vehicle Re-Identification

Is Better Teacher Supervision Enough? Unlocking Student-side Learning in Multimodal On-Policy Distillation

Is This Evidence Decision-Critical? Learning to Verify Rule-Governed Decisions

K2P: Label-Free Knowledge to Prompt Distillation

Marginal Response Surface Elicitation for Zero-Label Tabular Learning

More Choices, Fewer Decisions: Ordinal-Scale Bias in JEV-like Direct-Decision Models

Neuro-Symbolic Predicate Learning for Semantic Safe Robot Control

Structure vs. Chain-of-Thought: Evaluating LLM Criteria Extraction for Depression Severity

Uruqi: Learning Spatial Cognition from Visual Experience

When Clipping Reverses Correction: Failure Dynamics of Pointwise Forward-KL On-Policy Self-Distillation

CogWAM: Aligning Semantic Cognition with World Action Modeling via Event-Driven Interfaces

Correct Answers, Invalid Traces: What Verifiable Grade-School Math Reveals About Chain-of-Thought Traces

DatalogBench: Evaluating Large Language Models on Text-to-Datalog Synthesis

JEV-as-a-Judge: Accept When Confident, Escalate When Unsure

Learned Reporting Preferences in RLVR Can Conflict with the Current Request

RareDx: Controlled Knowledge Integration and Graph-Grounded Policy Optimization for Rare-Disease Diagnosis

RAWR: Reward Assignment Without Rollouts in Verifiable Domains

ReCAP: Retrieval-Guided Capability Reuse for Multimodal Continual Instruction Tuning

Reconstructing Implicit Scientific Knowledge: Evaluating LLM Agents through End-to-End Reproduction of Astronomy

Reliable but Design-Sensitive: Instrument Uncertainty in LLM Annotation

SERA: Scale-Equalized Rollout Allocation for Maximum Likelihood Reinforcement Learning

TAEC: Trajectory-Aware Evidence Coordination for Multi-Step Visual RAG

VideoLoop: Looped Working Memory Against Semantic Thrashing in Long-Form Video Agents

VISTA: Internalizing Collective Visual Experience via On-Policy Distillation for Active Multimodal Agents

VLALight: A Vision-Language-Action Model for Traffic Signal Control

What Shared Prefixes Hide: Trajectory Dropout for On-Policy Distillation

A Spectral Theory of Compositional Learning

ActionEngine: From Reactive to Programmatic Web Agents via State Machine Memory

Beyond the Training Horizon: Mechanisms and Limits of Length Generalization in Looped Transformers

Bring near-Astra intelligence to everyday work with GPT-6.1 Sol on Amazon Bedrock

Building Intelligent Agents with Neuro-Symbolic Concepts

Certified Interface Aliases: Exact Collisions in Vision-Language Preprocessing, and When They Exist

EEGAgentBench: Benchmarking LLM Agents on Short- and Long-Horizon EEG Analysis

Here are my 7 basic prompts after 4 year using experience

Learning Transferable Sensor Models via Language-Informed Pretraining

Making LLMs Truly Forget: Deep Unlearning by Searching, Selecting, and Severing Knowledge Paths

NavHarness: Towards Lifelong Embodied Navigation