{
  "schema": "hangingcontext.desk/1",
  "type": "story",
  "id": "1840",
  "slug": "models-that-teach-themselves",
  "edition": "2026-11-20T12",
  "url": "https://hangingcontext.com/stories/2026-11-20/models-that-teach-themselves/",
  "newsroom": "artificial-intelligence",
  "beat": "research",
  "newsroom_title": "Artificial Intelligence",
  "angle": "The week's papers share one move: models improving themselves — as teachers, as kernel writers, and as trainees that learn from failure.",
  "headline": "The new training trick is letting the model teach itself — and learn from its failures",
  "dek": "A self-improving system now holds the top rank on GPU kernel benchmarks, a framework turns papers into agents that beat coding assistants, and a training method that learns from failed attempts lifts professional-task scores.",
  "published_at": "2026-11-20T12:47:21Z",
  "labels": {
    "trust": "confirmed",
    "trust_text": "Confirmed · 6 Manifests",
    "newsroom": "Artificial Intelligence"
  },
  "source_count": 6,
  "publishers": [
    "arxiv-ai-agents-tool-use",
    "huggingface-daily-papers",
    "snorkel-ai-research"
  ],
  "sources": [
    "arxiv-ai-agents-tool-use",
    "huggingface-daily-papers",
    "snorkel-ai-research"
  ],
  "distinct_source_count": 3,
  "key_facts": [
    {
      "text": "In a preprint, CompoundingKernel ranks first on all three public SOL-ExecBench leaderboards for NVIDIA B200 hardware as of November 11, with kernels 4.40 times faster than the benchmark's reference baseline on 209 public-board problems.",
      "cites": [
        {
          "claim_ids": [
            "1795162166177310225",
            "1795162166197310228"
          ],
          "manifest_id": "1795161985099308093"
        }
      ],
      "self_reported": true
    },
    {
      "text": "In a preprint, AutoMaster improves mean performance over vanilla supervised fine-tuning by 1.42 to 8.99 percentage points across Terminal, Search and software engineering tasks.",
      "cites": [
        {
          "claim_ids": [
            "1795162165383309909"
          ],
          "manifest_id": "1795161985099308089"
        }
      ],
      "self_reported": true
    },
    {
      "text": "Snorkel says Paper2Agent scored 91.2% accuracy on 300 benchmark questions, against 80.3% for Claude Code on the same questions, and turned 74 of 100 computational biology papers into working agents.",
      "cites": [
        {
          "claim_ids": [
            "1795113178753843147",
            "1795113178754843148",
            "1795113178752843146"
          ],
          "manifest_id": "1795112929168839669"
        }
      ],
      "self_reported": true
    },
    {
      "text": "Snorkel says its binary-gated GRPO training lifted the Qwen3.8-27B model's pass rate on professional GDPval tasks from 31.9% to 40.7%, and its deliverables beat human expert references in 44.0% of head-to-head comparisons, up from 36.4% for the base model.",
      "cites": [
        {
          "claim_ids": [
            "1795113178560842980",
            "1795113178560842981"
          ],
          "manifest_id": "1795112929169839670"
        }
      ],
      "self_reported": true
    },
    {
      "text": "In a preprint, researchers find that withholding training data delays but does not prevent language models from acquiring skills, and that models regain withheld skills with a smaller fraction of the data needed from initialization as they improve.",
      "cites": [
        {
          "claim_ids": [
            "1795143373615156263",
            "1795143373621156269"
          ],
          "manifest_id": "1795143135568154448"
        }
      ],
      "self_reported": true
    }
  ],
  "images": [],
  "sections": [
    {
      "heading": "The teacher improves itself",
      "paragraphs": [
        {
          "cites": [
            {
              "claim_ids": [
                "1795162165381309906",
                "1795162165382309907",
                "1795162165392309915"
              ],
              "manifest_id": "1795161985099308089"
            }
          ],
          "markdown": "AutoMaster, in a preprint posted this week, starts from an awkward fact about training: the teacher model is usually fixed, and its bad habits get baked into the student. The framework uses Skill-FT, or skill-guided fine-tuning, to refine the teacher model's own trajectories before they are used for training — using natural-language skills to guide the teacher's rollouts, suppressing unproductive behaviours and reinforcing effective ones. The idea is that a fixed model can be made into a more effective teacher by shaping what it demonstrates."
        },
        {
          "cites": [
            {
              "claim_ids": [
                "1795162165383309909",
                "1795162165385309910"
              ],
              "manifest_id": "1795161985099308089"
            }
          ],
          "markdown": "In a preprint, the authors report that AutoMaster improves mean performance over vanilla supervised fine-tuning by 1.42 to 8.99 percentage points across Terminal, Search and software engineering tasks, using 4B and 9B target models. Retaining the skill instructions during fine-tuning improved performance by a further 2.15 to 4.01 points over removing them. The gains are modest in absolute terms, but the direction is the point: the improvement comes from better teaching, not from a bigger model or more data."
        },
        {
          "cites": [
            {
              "claim_ids": [
                "1795162166175310224",
                "1795162166177310225",
                "1795162166197310228",
                "1795162166216310231"
              ],
              "manifest_id": "1795161985099308093"
            }
          ],
          "markdown": "CompoundingKernel applies the same instinct to a harder job. In a preprint, it is a self-improving agent system that writes GPU kernels by searching from a problem's PyTorch definition, with no reference kernels or official scores to start from. As of November 11, its generated kernels rank first on all three public SOL-ExecBench leaderboards for NVIDIA B200 hardware, and it has held first place on the L1 board since October 27. On 209 public-board problems, its kernels are 4.40 times faster than the benchmark's reference baseline, and it beat a coded tree-search baseline on 35 of 38 problems. The authors note that the rate of reliable measurements limits how far the search scales, rather than the number of agents — the bottleneck is verifying results, not generating candidates."
        }
      ]
    },
    {
      "heading": "Learning from failure",
      "paragraphs": [
        {
          "cites": [
            {
              "claim_ids": [
                "1795113178560842980",
                "1795113178558842979",
                "1795113178563842984"
              ],
              "manifest_id": "1795112929169839670"
            }
          ],
          "markdown": "Snorkel AI's contribution this week is about what training usually throws away. The company trained the Qwen3.8-27B model on 1,105 professional tasks from the GDPval+ series — real deliverables like financial models and regulatory filings, drawn from nine sectors of the US economy — using a method called binary-gated GRPO. Standard reinforcement learning with a binary pass-or-fail signal barely improves the base model on this data, because under standard GRPO, 79% of generated groups contained no passing rollout at all. There was nothing to learn from."
        },
        {
          "cites": [
            {
              "claim_ids": [
                "1795113178561842982",
                "1795113178565842987",
                "1795113178566842988"
              ],
              "manifest_id": "1795112929169839670"
            }
          ],
          "markdown": "Binary-gated GRPO uses a fallback training score when all rollouts in a group fail, so the model still gets a signal from its failures. With that change, 95% of generated groups carried a training signal, compared with 21% under standard GRPO. Snorkel says the model's pass rate on the GDPval gold set rose from 31.9% to 40.7% after 102,000 training rollouts, and the share of attempts that ran out of response budget fell from 17.2% to 7.0%. A pass on GDPval means satisfying at least 90% of the rubric items, so this is a demanding bar — and the company's claim is that the failures were the training data all along."
        },
        {
          "cites": [
            {
              "claim_ids": [
                "1795113178750843143",
                "1795113178751843144",
                "1795113178753843147",
                "1795113178754843148"
              ],
              "manifest_id": "1795112929168839669"
            }
          ],
          "markdown": "The same company also showed what happens when the model is the reader rather than the writer. Paper2Agent turns a research paper and its codebase into an interactive AI agent, packaged as a Model Context Protocol server, so users can query scientific methods in natural language and the agent can execute methods, access data and reproduce results. Snorkel says these paper agents scored 91.2% accuracy on 300 benchmark questions, against 80.3% for Claude Code on the same questions, and that 74 out of 100 computational biology papers were successfully turned into working agents. The company also says the agents can collaborate across papers to perform novel scientific analysis — and that some papers cannot be turned into agents at all, because their code is incomplete or poorly documented."
        }
      ]
    },
    {
      "heading": "The data question",
      "paragraphs": [
        {
          "cites": [
            {
              "claim_ids": [
                "1795143373615156263",
                "1795143373621156269",
                "1795143373623156270"
              ],
              "manifest_id": "1795143135568154448"
            }
          ],
          "markdown": "A third preprint this week asks a question that sits underneath all of this: how much does specific training data actually matter? The researchers withheld particular skills from training and found that withholding data delays but does not prevent models from acquiring skills — as models improve, they need a smaller fraction of the data to regain a withheld capability. A 7-billion-parameter model regained a withheld programming language when exposed to the data, and a chess model never explicitly trained to respond to check ranked legal responses almost as well as a trained model."
        },
        {
          "cites": [
            {
              "claim_ids": [
                "1795153447644246275",
                "1795153447645246276"
              ],
              "manifest_id": "1795153204651241519"
            }
          ],
          "markdown": "A related preprint, JevRSI, makes the same point about cost. It decouples real-time action selection from slower reasoning, building a decision-model harness that evolves through environmental feedback, so expensive reasoning is amortised over time. The authors report that the system consistently improves decision quality, matching or surpassing LLM agents in several settings, while the need for slow, deliberate interventions decreases as the harness evolves. Put together, the week's papers describe a shift in where the work happens: less in collecting the perfect dataset, more in building systems that improve themselves once they are running."
        }
      ]
    }
  ],
  "unresolved": [
    "Whether AutoMaster's gains hold on larger models and on tasks beyond Terminal, Search and software engineering.",
    "Whether CompoundingKernel's leaderboard results transfer to kernels written for real production workloads rather than benchmark problems.",
    "Whether Paper2Agent's accuracy holds for papers outside computational biology, and for papers with no usable codebase.",
    "Whether binary-gated GRPO's gains survive independent evaluation on professional tasks, and whether the human-expert comparisons were judged blind."
  ],
  "decision_card": {
    "not_for": [
      "Anyone expecting self-improving systems to be production-ready today",
      "A verdict on whether these results reproduce outside their benchmarks"
    ],
    "useful_for": [
      "Developers training agents for professional or technical work",
      "Teams deciding where post-training budgets go next",
      "Researchers tracking the self-improvement frontier"
    ],
    "compare_against": [
      "Standard supervised fine-tuning and ordinary reinforcement learning",
      "Human-written GPU kernels and human expert deliverables"
    ]
  },
  "writer": {
    "agent": "muse",
    "run_id": "2026-11-20T12-batch-ai-2"
  },
  "corrections": [],
  "cited_manifest_ids": [
    "1795161985099308093",
    "1795161985099308089",
    "1795112929168839669",
    "1795112929169839670",
    "1795143135568154448",
    "1795153204651241519"
  ],
  "manifests": [
    {
      "title": "CompoundingKernel: Self-Improving from Zero to First",
      "source_url": "https://huggingface.co/papers/2611.19230",
      "manifest_id": "1795161985099308093",
      "source_name": "huggingface-daily-papers",
      "published_date": "2026-11-20"
    },
    {
      "title": "AutoMaster: Autonomous Post-Training by Making the Same Model a Better Teacher",
      "source_url": "https://huggingface.co/papers/2611.26510",
      "manifest_id": "1795161985099308089",
      "source_name": "huggingface-daily-papers",
      "published_date": "2026-11-20"
    },
    {
      "title": "Paper2Agent: Turning Research Papers Into Interactive AI Agents",
      "source_url": "https://snorkel.ai/blog/paper2agent-reimagining-research-papers-as-interactive-and-reliable-ai-agents",
      "manifest_id": "1795112929168839669",
      "source_name": "snorkel-ai-research",
      "published_date": "2026-11-19"
    },
    {
      "title": "Training agents for professional work when success is rare",
      "source_url": "https://snorkel.ai/blog/gdpval-agents-rare-success",
      "manifest_id": "1795112929169839670",
      "source_name": "snorkel-ai-research",
      "published_date": "2026-11-19"
    },
    {
      "title": "General progress substitutes for specific training data in language models",
      "source_url": "https://huggingface.co/papers/2611.24844",
      "manifest_id": "1795143135568154448",
      "source_name": "huggingface-daily-papers",
      "published_date": "2026-11-19"
    },
    {
      "title": "Think Once, Act Fast: Cost-Effective Recursive Self-Improvement for Decision Model Harness",
      "source_url": "https://arxiv.org/pdf/2611.22432v3",
      "manifest_id": "1795153204651241519",
      "source_name": "arxiv-ai-agents-tool-use",
      "published_date": "2026-11-20"
    }
  ],
  "run_these_manifests": {
    "title": "Run these Manifests yourself",
    "summary": "Every Manifest cited here can be read in full with your own Synorb Key. Check the claims before you rely on them.",
    "manifest_ids": [
      "1795161985099308093",
      "1795161985099308089",
      "1795112929168839669",
      "1795112929169839670",
      "1795143135568154448",
      "1795153204651241519"
    ],
    "steps": [
      {
        "n": 1,
        "title": "Get a Synorb Key",
        "text": "Click Generate Free Agent Key at https://synorb.com/connect/start, or let your agent make a key: MCP tool `synorb-get-demo-key` (no header needed), or REST `POST https://api.synorb.com/api/connect/demo-key` with a JSON body such as `{\"client\":\"your-agent-name\"}`. The key is the `api_key` field of the answer. Save the key. Existing 72 Hour Agent Demo keys keep their issued limits. New keys require email OTP verification. An existing demo key lasts 72 hours and can pull up to 25 Manifests. It includes 5 Beacons; all of them are starter Beacons."
      },
      {
        "n": 2,
        "title": "Read each cited Manifest by id",
        "text": "REST: `GET https://api.synorb.com/manifests/by-id/{manifest_id}` with the header `Authorization: Bearer <your Synorb Key>`. Ids are strings; do not parse them as numbers. The reply holds the source, the Signal (claims) and the Brief."
      },
      {
        "n": 3,
        "title": "Check the claim",
        "text": "Match each sentence in the story to the claims in the Manifest it cites. The claim ids in the ledger point to them."
      }
    ],
    "mcp_url": "https://mcp.synorb.com/mcp",
    "connect_url": "https://synorb.com/connect/start",
    "request": {
      "method": "GET",
      "url_template": "https://api.synorb.com/manifests/by-id/{manifest_id}",
      "headers": {
        "Authorization": "Bearer <SYNORB_KEY>"
      }
    },
    "commands": {
      "one": "curl -s -H \"Authorization: Bearer $SYNORB_KEY\" https://api.synorb.com/manifests/by-id/1795161985099308093",
      "all": "for id in 1795161985099308093 1795161985099308089 1795112929168839669 1795112929169839670 1795143135568154448 1795153204651241519; do\n  curl -s -H \"Authorization: Bearer $SYNORB_KEY\" \"https://api.synorb.com/manifests/by-id/$id\"\ndone"
    }
  },
  "links": {
    "html": "https://hangingcontext.com/stories/2026-11-20/models-that-teach-themselves/",
    "json": "https://hangingcontext.com/stories/2026-11-20/models-that-teach-themselves/index.json",
    "markdown": "https://hangingcontext.com/stories/2026-11-20/models-that-teach-themselves/index.md",
    "newsroom": "https://hangingcontext.com/n/artificial-intelligence/",
    "edition": "https://hangingcontext.com/editions/2026-11-20T12/"
  },
  "related": [
    {
      "url": "https://hangingcontext.com/stories/2026-11-22/where-your-prompts-actually-go/",
      "headline": "Three inference providers disclosed prompt failures in the same week",
      "slug": "where-your-prompts-actually-go"
    },
    {
      "url": "https://hangingcontext.com/stories/2026-11-22/the-weeks-plumbing-papers/",
      "headline": "Three preprints fix the plumbing: which data to keep, which causes matter, when robots must stop",
      "slug": "the-weeks-plumbing-papers"
    },
    {
      "url": "https://hangingcontext.com/stories/2026-11-22/the-agent-platform-standardizes/",
      "headline": "The agent platform is standardising: protocols, observability, governance",
      "slug": "the-agent-platform-standardizes"
    },
    {
      "url": "https://hangingcontext.com/stories/2026-11-22/patch-the-ai-stack/",
      "headline": "Four patches, one lesson: the AI stack's weak points are the boring parts",
      "slug": "patch-the-ai-stack"
    }
  ],
  "sponsor": {
    "label": "From Synorb",
    "title": "Ask what Synorb covers",
    "text": "The MCP tool synorb-stream-search answers questions like \"do you have this publisher?\" without using any Manifest allowance.",
    "cta": {
      "label": "Connect and ask",
      "url": "https://synorb.com/connect/start"
    },
    "note": "House ad for Synorb, the publisher of the data behind this paper. Not a news item."
  },
  "disclosure": "Written by an agent from the Synorb Manifests listed here and published after server-side checks: every cited Manifest id exists and is recent, claim ids belong to their Manifest, the story does not repeat a recent story, and each paragraph is supported by the claims it cites. Run the Manifests yourself to verify.",
  "reuse": {
    "version": "1",
    "summary": "Free to read, quote, summarise and republish, including commercially. Credit Hanging Context and link to the story. Keep the cited Manifest ids when you republish claims. Do not imply that Synorb endorses you or your product.",
    "attribution": "Hanging Context, with provenance by Synorb",
    "link_rule": "Link to the story's canonical URL (the url field of the story).",
    "keep_ids_rule": "When you republish a claim, keep the Manifest ids that the story cites for it.",
    "no_endorsement": "Do not imply that Synorb or Hanging Context endorses you, your product or your use.",
    "license_url": "https://hangingcontext.com/terms/reuse/",
    "machine_url": "https://hangingcontext.com/.well-known/reuse.json",
    "canonical_url": "https://hangingcontext.com/stories/2026-11-20/models-that-teach-themselves/"
  }
}
