WRITER · Research Lab

Frontier research for enterprise AI.

Long context, self-improving models, agent evaluation and reliability in finance and medicine. Every paper here comes with an explanation of how it works and an interactive demo you can use.

Publications
27
Interactive
27
Since
2020
  1. PaperMulti-layer Indexer+14.5%decode speed over IndexCache on DeepSeek-V3.2 (retention 1/2)LCFM Workshop, NeurIPS 2026 →
  2. PaperDenoising blocks, not tokens>80×fewer generation FLOPs than the ELF-L diffusion baselineInteractive →
  3. ModelPalmyra x60.785BFCL Core, the highest in its comparison cohortInteractive →
  4. PaperThe Harness Effect−41%blended cost per task ($0.21 → $0.12), quality at parityInteractive →

FIG. 01: Mechanism

Move through the papers. The tokens rearrange into each one’s mechanism.

Featured research

Inside the research

Eight papers to start with. Every paper in the archive has its own interactive demo.

  1. 01Reflect, Retry, RewardWhen a model fails, it writes a reflection on what went wrong and tries again. If the retry succeeds, only the reflection tokens get rewarded. With nothing more than a binary pass/fail signal, small models learn to self-correct well enough to beat models ten times their size.Paper · May 30, 2025+34.7%math equation writing (Countdown)
  2. 02Writing in the MarginsLong documents are processed in chunks, and the model writes short notes in the margin as it reads each one. The relevant notes are carried into the final answer. Off-the-shelf models handle long context better, with no fine-tuning.Paper · Aug 27, 2024+7.5%average accuracy on multi-hop reasoning (HotpotQA, MultiHop-RAG)
  3. 03FailSafeQAReal users misspell, ask half a question and upload the wrong document. FailSafeQA stress-tests LLMs on 10-K filings under these conditions and measures two things: whether a model stays accurate, and whether it admits when it can't answer.Benchmark · Feb 10, 20250.80best Context Grounding (Palmyra-Fin-128k-Instruct)
  4. 04Sycophancy & memoryMemory systems are supposed to make assistants better. We found they can make them agree with you even when you're wrong. When a memory system extracts facts, it often drops the correction the assistant made earlier, and the model then sides with the user's mistaken belief.Paper · Jun 9, 20261.6% → 40.2%Sonnet 4.6 sycophancy on MIST-Moral: chat history vs. Mem0 memory
  5. 05Prediction ≠ preventionA critic that predicts agent failures with 0.94 AUROC sounds like a safety net. Put it in charge of intervening, though, and it can cut success by 26 points. Intervening rescues some failing runs and breaks some that would have succeeded. We show how to measure that trade-off before you deploy.Paper · Feb 3, 20260.94critic AUROC (an accurate failure predictor)
  6. 06Outcome-oriented agent evaluationLatency and time-to-first-token say how fast an agent is. They say nothing about whether it did the job. This white paper proposes eleven outcome-based metrics, including goal completion, autonomy, resilience and business impact, and uses them to compare four agent architectures across five industries.Paper · Nov 11, 202511outcome-based, task-agnostic metrics
  7. 07OmniACTCan an AI agent look at a screen and write the script that completes a task, from "play the next song" to sending an email? OmniACT is a benchmark of desktop and web tasks for answering that. The strongest baseline, GPT-4, reached only 15% of human proficiency.Benchmark · Feb 27, 202415%of human proficiency reached by the strongest baseline (GPT-4)
  8. 08The Harness EffectAgent bills grow with every reasoning trace, tool schema and sub-agent prompt. Holding six models fixed and swapping only the orchestration layer, the Writer Agent Harness cut cost per task 41%, wall-clock time 44% and tokens 38%, with quality at parity.Paper · Jul 8, 2026−41%blended cost per task ($0.21 → $0.12), quality at parity

FIG. 03: Program

Four directions, one standard: does it hold up in production?

  1. Enterprise-optimized models

    Models built for reliability, long context, tool use and cost at enterprise scale.

  2. Practical evaluations

    Evaluation methods that reflect real deployments, real risks and real failures.

  3. Domain-specific specialization

    Specialized models and benchmarks for finance, medicine and other high-stakes domains.

  4. Retrieval & knowledge

    Retrieval and knowledge integration that holds up on real enterprise data.

FIG. 04: Archive

All research