NAME THAT LLM / TERMINAL

The public 0.683 recipe run unchanged on our account. The author reported 0.683 and we got 0.679, 0.003 under our best.

The public 0.683 recipe with our new full-pool Qwen model mixed in at ten percent. It lost to the same recipe without it.

The public 0.682 recipe with ten percent tokenizer models and twenty percent raw-byte models mixed in. It lost.

The public 0.682 recipe run unchanged on our account. It matched the author's score and is our best entry.

The public 0.682 recipe with ten percent of our earlier Qwen3-8B model mixed in. It lost.

The public 0.682 recipe with ten percent of our tokenizer model mixed in. It lost.

A full run of the public 0.681 ensemble on our account. It beat our old best of 0.651 by 0.030.

Our own trained Qwen3-8B model submitted by itself to test the pipeline end to end. It scored 0.585, close to its 0.588 local score and well below the public ensemble.

The public neural ensemble, the 0.651 recipe, run in full on our account. Our first strong board score.

Our own character, word and style classifier, trained on the official labels, to prove an entry of ours runs on the hidden test.

A second try at the public 0.651 notebook, still with our extra wrapper code. It crashed on the hidden test and earned no score.

The first try at the public 0.651 notebook, with our extra wrapper code. It crashed on the hidden test and earned no score.

Gaurav Rawat asks why Qwen classifies so well. The saved file has only the title and no replies.

somerandomguy says the time limit counts from kernel start, not queue, use the competition-rerun flag to detect scoring runs, and keep about half an hour spare. Check our notebooks.

Only the title was saved: pedrohbenetton's notes on fold splits by prompt, prompt-group tricks that did not carry over to test, and punctuation clues. Body missing; fetch it before relying on it.

Only reply: Darragh1Walsh links a dataset he made with Kaggle Benchmarks model access and invites use. A free extra data source, but check it matches the host format first.

Maxim Stepanov asks if a CatBoost approach can compete; the only reply is 'wondering the same'. No answer.

Vijay Konda says GLM-5.3 likelihood only fits the time limit by streaming one layer at a time over all test passages; John K shows GLM and Claude labels heavily confused. Worth a test.

Title only (James Bedichek, model parameterizations), no body or replies saved. Nothing to act on.

Under-18 consent procedure question, no replies. Not relevant to us.

Renan Pinheiro says random folds score 0.11 but folds grouped by prompt score 0.35 (board 0.355); most errors stay in the right family. Use grouped folds and try a family-then-model step.

subuta_umee asks if outside data made by sending competition text to other LLM services is allowed; no official answer, one unhelpful reply. Rule still unclear.

Emmanuel Ocheme offers Mistral and OLMo LoRAs (public 0.663). Anto Pil says they add only 0.004 to his 0.666 blend and declines to merge. His and Emmanuel's fold predictions are public; test them in our blend.

Empty in our copy: title promises per-part public LB ablations on two bases (0.638 vs 0.659), but no body or comments were saved. Worth fetching the post itself.

radidani's post on prize winner requirements. The saved file has only the title, no body or replies.

Matin shirazi shares a style baseline with grouped cross-validation notebook; no comments. Our style models already cover this.

Title only (Mohammed Hasan Abdullah, public 0.333), no body or replies saved. Nothing useful for us.

John K says train has 18 answers on 149 of 150 prompts, but test groups are not full and model counts differ. Never force one-of-each-model per prompt at test time.

cylovelife's write-up on grouped validation, style baselines and failed neural runs. The saved file has only the title, no body or replies. Nothing to act on.

Newcomer asks why test has 3 rows; somerandomguy says that is a sample and the hidden test has about 4000 rows. Known already.

Qhaleesh posts a DistilBERT baseline at 0.25. OsakanaTeishoku's only reply says to split folds by prompt, since test prompts are unseen. We already do that; nothing new for us.

somerandomguy says his self-generated data lowered val even after matching structure and length; Soham Acharya fears topic bias. Supports not paying for our own generation.

AyoGG says their classifier calls GLM-5.3 Claude; commenters blame distillation or routing to Claude. This confusion pair is a likely error spot to check in our results.

Joke thread about LLM-written posts on the forum; one user says he is stuck around 0.66. Nothing useful for us.

Best single models: Horikita Saku and Xiao say 0.65 to 0.68 with Qwen3-14B and own data; GoPain says Qwen3-8B LoRA 0.628, ensemble adds under a point. Most gains came from data, so more clean data beats more blending.

Comments are just a team-up request (bosprimigenious, LB 0.349) and an upvote. Nothing about whether the 9,719 OpenAI-response dataset helps.

Carrot Production says a quarter of the public pool was junk or wrong genre, and cleaning it (Open Pool v9) gave +0.027. Nobody has replied. Try that cleaning on our full-pool Qwen training.

Harshita: swapping old Qwen2.5-14B for a higher-score retrain cost board points; HALX2158 says diverse weak members matter. GoPain: centred-head weight over 1.0 lost. Add diversity, not clones.

Carrot Production updates the Open Pool to v14, with Kimi relabelled and wrong rows removed. GoPain shows a board gap of 0.002 between two models is noise. Train on v14 and judge gains by our own cross-validation.

Only answer: Cloud9 links an arXiv paper (ATLAS) that predicts the provider well. No one reports results for provider-then-model here. Nothing to act on.

Adisak Sukul asks six questions on training data made from the candidate models; rw-nue asks about a spending cap and release deadline. No host answer, so the rule stays unclear.

HALX2158 says Gemini never em-dashes in train but proxy-made Gemini text does (13-43%); Nikhil sees Claude and GLM confused. Punctuation rules may flip if we add outside Gemini data.

Aleksey Osipenko says early Qwen layers predict the author better than late ones, and top public work cuts Qwen3-8B to 24 layers and pools several. About half the errors are within one model family.

GoPain says grouped CV reads 0.03-0.04 high vs LB once the pool is in training, and class balancing gave +0.011 on LB. Our pool-trained local gains are inflated; keep balancing.

prvsiyan shares a helper dataset; Darragh1Walsh adds one he updates daily. Commenters doubt its labels because it came through an unofficial proxy. Use only if checked against official data.

Zbigniew Kirchner says extra Claude passages in a side Claude head on a Qwen3-8B LoRA lifted val 0.456 to 0.493, and a lab-then-version hierarchy gave +2.4. Cheap Claude-split idea to try.

Yan Zhou asks the hosts for more official training data. buysider asks whether self-generated text from model providers is even allowed; Yan doubts hosts will answer. No ruling given.

Oleg Zholobov built the 18-model corpus from Hugging Face and GitHub eval logs. HALX2158 says its Markdown and line breaks hurt CV even flattened. Filter to short single paragraphs.

Efe Karamese asked if self-made API data is allowed; haritch asks whether public API-made datasets on Kaggle are fine. No host answer. Unresolved rule risk if we use those datasets.

YePing Lin asks whether anyone tried the PersonalLLM dataset; no replies. Just an open question, nothing to act on.

Only the title was saved: pjmathematician says he rebuilt the train prompts for all 150 prompt groups and shared them as a dataset. No body or replies. Worth a look as a source for new training text.

HALX2158 says the public 18-model corpus lowered grouped CV (format mismatch). James Warner puts extra data at about +0.11. Only data shaped like the host's helps us.