Fetching from the wire…
01
02
03
04
05
06
07
08
Skills2026-07-30 · source-backed
Fragments from the same function or class currently eat multiple context slots with redundant renderings, crowding out genuinely different code. Canonicalize per code object, admit at most one companion fragment and only when it carries semantics the canonical view lacks, then render under both per-object and global token budgets. VITAL-RAG lifted RepoBench Recall@4K from 39.59% to 63.67% while cutting evidence tokens 35.63%.
Each link below shares sources, entities, or timing with this story.
arXiv 2608.00765 compresses retrieved docs into query-conditioned visual representations, sidestepping the trade-off where hard compression is query-aware but weak and soft compression is strong but needs costly offline encoding. Beats both baselines across varying retrieval d...
1. Flip your multi-model pipeline to review-then-generate. Instead of using a reasoning model to plan before code generation, let the specialist generate freely and use reasoning tokens for review. Paper shows 90.2% pass@1 vs 87.2% for the planning pattern. Source 2. Audit you...
arXiv 2607.26805 uses a lightweight draft model to generate drafts over partitioned repository contexts, then Structure-Aware Draft-Guided Selection picks informative context by API consistency and logical similarity before a reduce phase aggregates, with parallel verification...
Every retrieval pipeline I've built follows the same instinct: rank, threshold, pass only the top hits. Noise is bad. Precision is good. A controlled study says that instinct costs you accuracy (arXiv 2608.17188). 2,420 trials, 11 model configurations, 661 anonymized workplace...
Rust's ownership and borrowing rules trip LLMs into rule-violating or unsafe code, so AdaTrans (arXiv 2606.31706) maps specific compiler errors to targeted repairs via strategy-driven RAG and error-stratified transformation, cutting reliance on unsafe constructs. If you're doi...
A new reliability protocol (arXiv:2606.07783) tests RAG on factoid questions the model already answers correctly without retrieval, then injects clean, misleading, and mixed evidence to measure when retrieved content overrides correct internal knowledge. It introduces a parame...
MindPattern daily
One email a day at 7 AM. Sources and a take on every story. Unsubscribe anytime.