SYNTHESIS NOTE
Topics›Flaws›this note

Do frontier LLMs silently corrupt documents in long workflows?

DELEGATE-52 tests whether state-of-the-art language models reliably preserve document integrity across extended delegated tasks. Understanding this matters because single-step benchmarks may mask compounding failures that emerge only at workflow scale.

Synthesis note · 2026-05-18 · sourced from Flaws

Delegation requires trust — the expectation that an LLM will execute a task without introducing errors. DELEGATE-52 stress-tests that expectation with 310 work environments across 52 domains (coding, crystallography, music notation, genealogy) and a round-trip relay protocol where each task is paired with its inverse, so a perfect model would recover the original document exactly.

Across 19 LLMs, even frontier systems (Gemini 3.1 Pro, Claude 4.6 Opus, GPT 5.4) corrupt an average of 25% of document content by the end of long workflows. Weaker models fail more severely. The degradation curve decelerates but does not plateau — the first half of an extended relay accounts for 2-3x more loss than the second half, yet the strongest model still drops below 60% accuracy by round-trip 50. Distractor files, longer documents, and longer interactions all worsen the rate.

The structural problem: errors are sparse but severe and they compound silently. A user reviewing one or two outputs sees competent work. A user delegating an end-to-end workflow gets a document that looks intact but contains accumulated drift in places they did not check. The trust assumption that holds at single-step interaction collapses at the timescale where delegation is actually valuable.

This is not a "weak model" finding. It is a ceiling on delegated work at the current frontier — one that scales unfavorably with exactly the workflow length that makes delegation attractive.

Inquiring lines that read this note 157

This note is a source for these research framings, grouped by the broader line of inquiry each explores. Scan the bold lines of inquiry; follow any specific question forward.

How does tokenizing intelligence reshape labor markets and institutional outcomes? What explains the gap between LLM semantic understanding and production? Can safety evaluations accurately measure true capabilities when models obscure them? When should retrieval be triggered to augment language model reasoning? How do frontier models develop behaviors that escape detection? How can model evaluations remain robust against gaming and evaluator drift? What reasoning limitations and biases do language models systematically exhibit? Do AI safety evaluations reliably detect unsafe model behavior? Why does separating planning from execution enhance reasoning performance? How do different editing strategies affect output quality and agent safety? Do language models truly understand their encoded knowledge or merely retrieve patterns? Why do autonomous agents fail and how well can they detect failures? How does policy entropy collapse limit reasoning RL scaling? Does parallel reasoning outperform sequential approaches in reasoning efficiency? How should context management strategies balance efficiency with agent reliability? How do dialogue structure and state management affect multi-turn reasoning quality? How can knowledge graph structure improve reasoning and retrieval? How does explicit structure improve LLM reasoning on complex tasks? What systematic failures occur in human preference-based model evaluation? How does verifier integration affect the generation-verification gap? How do different post-training structures affect learning outcomes? Can language models develop genuine communicative grounding with humans? Can synthetic data preserve necessary diversity to prevent model collapse? Does reinforcement learning create new reasoning capabilities or refine existing ones? Can verification capability keep pace with AI generation speed? What systematic biases do LLM judges exhibit in evaluation? How should conversational agents balance proactive goal-pursuit with user autonomy? Can we reliably distinguish AI-generated from human-written text? What can prompting techniques actually accomplish in language models? How do benchmark scores relate to real-world task completion? Can models strategically underperform to evade capability evaluation? Does AI-assisted coding actually improve developer productivity and code quality? How do standardized forms and measurement artifacts shape observable evidence? How should organizational structure guide multi-agent system design? Does chain-of-thought reasoning faithfully reflect actual model reasoning processes? Should agent evaluation measure execution quality beyond task success? How do standardized artifacts improve coordination in multi-agent systems? Why does consolidated memory sometimes hurt continual learning more than no memory? How can outcome-only reporting obscure the true mechanisms of safety? How do workflow position and structure enable attacks in multi-agent systems? When does governance become the bottleneck for agentic AI? What determines whether visible reasoning traces improve task performance? Can post-hoc execution traces provide sufficient accountability without real-time oversight? How can agents learn and retain useful knowledge across diverse environments? What prevents or enables error accumulation in self-training systems? How does model scale affect fine-tuning and downstream task performance? How should agent memory systems balance granularity, management, and reuse? Is model misalignment inherent to base models or induced by alignment procedures? What external mechanisms enable reliable self-improvement in language models? What is the actual bottleneck preventing effective AI research automation? Does polished AI output mislead users about actual accuracy and expertise? Can validator agreement alone ensure semantic correctness of reasoning? How can language models effectively manage long-context memory without quality degradation? Do standard benchmarks reveal AI capability in real-world contexts? What defense strategies address attacks that per-skill scanning cannot detect? What are the trade-offs between memory capacity and continuous task learning? What makes reasoning models struggle with ill-posed and unfamiliar problems? Does architectural recurrence enable computational capabilities that fixed-depth models cannot achieve? Which agent-created code should persist in infrastructure versus remain task-specific? Can formal safety constraints prevent AI agents from exceeding authorized boundaries? Why doesn't LLM knowledge translate to improved reasoning outputs? How should control and autonomy be allocated in human-AI collaboration? Can we reliably measure and demonstrate AI's actual productivity impact? How can we better validate AI research and prevent misconceptions? Do single-axis benchmarks adequately measure deployment-ready agent capability?

Related concepts in this collection 6

This note in its neighbourhood — explore the map, then jump to a related concept in the list below.

Concept map
15 direct connections · 147 in 2-hop network ·dense cluster Open in graph ↗
Do frontier LLMs silently corrupt documents in l… Does model capability change how documents degrade… Can better tools fix LLM document editing errors? Do short benchmarks predict how models perform ove… Do models fail worse when their own errors fill th… Why do language models fail to act on their own re… Can safety tests miss hazards that build over time…

Click a node to walk · click center to open · click Open in graph to see this note in the full knowledge graph

your link semantically near linked from elsewhere

Related papers in this collection 8

Papers most semantically related to this note, ranked by cosine similarity in the embedding space.

Search by related questions 5

Suggested questions this note speaks to — click to search the collection, or type your own.

Why do frontier models corrupt documents while weaker models delete them?Can tool use or self-conditioning fix long-horizon delegation drift in LLMs?Do short interaction benchmarks predict how LLMs perform in long workflows?Why do sparse per-step errors accumulate undetected across delegated tasks?How can static safety tests miss risks that emerge over time?
Original note title

frontier LLMs silently corrupt 25 percent of document content over long delegated workflows without plateauing