Open peer review, then a validated version of record.
Request peer review for your work. Independent reviewers assess it openly, their reports are published with the article, and the outcome follows their assessment of soundness.
Request review. Get open reports. Publish a validated version.
Your work is public from day one. Reviewers assess it in the open, and the article is labelled Peer reviewed when their reports say it is sound. There are no editors and no desk rejections for novelty.
01
Request
Day 0
Choose Publishing when you submit, or request review for a preprint or archive record you have already posted.
02
Screening
Within 3 days
A short administrative check for scope, authorship, licence and files. No judgement on novelty or impact.
03
Open review
Target 28 days
At least three independent reviewers are invited. Each one publishes a report with one of three outcomes.
04
Respond & revise
At your pace
Answer every report publicly and post a new version. Reviewers can update their assessment.
05
Outcome
On approval
The article is labelled Peer reviewed once reviewers approve it. No editor decides; the reports do.
How outcomes are determined
Peer reviewed
Two Approved reports, or one Approved and two Approved with reservations.
Under peer review
Reviewers invited or reports still coming in. Everything so far is public.
Not approved
Two Not approved reports. The article stays online with its reports.
Each report ends in Approved, Approved with reservations or Not approved. Read the full peer-review policy.
Reviewers assess
Soundness of methods and analysis
Whether conclusions follow from the evidence
Reproducibility: code, data, variance
Clarity and completeness of reporting
Ethics and competing interests
Reviewers do not
Novelty or originality
Expected impact or citations
Fit with a journal's “scope”
Whether results are positive
Already posted a preprint?
Request peer review for any record you own. Your preprint keeps its DOI and gains a peer-reviewed version.
Version 1 · 0 reviewer reports · 5 October 2026 · 10.xxxx/ctm.pb.2026.0009
Users frequently mix languages within a single query, but retrieval benchmarks rarely reflect this. We construct code-mixed query sets for two language pairs and evaluate dense retrievers under different tokenisation and fine-tuning strategies. Adapting the query encoder with synthetic code-mixed data improves recall@20 by 9 points over a multilingual baseline, while transliteration normalisation helps for romanised queries but hurts for native-script queries.
Version 3 · 3 reviewer reports · 4 September 2026 · 10.xxxx/ctm.pb.2026.0005
Debate about autonomous AI agents often centres on what they can do. This perspective argues that governance should instead start from who is accountable when an agent acts. Drawing on analogies from aviation, finance and medical devices, we propose three design requirements for deployed agents — attributable identity, bounded authority and reconstructable action logs — and discuss how each could be audited. We argue that these requirements are technically modest, can be adopted before agents become more capable, and would make later regulation easier to apply.
Version 1 · 1 reviewer report · 28 August 2026 · 10.xxxx/ctm.pb.2026.0006
Running assistant-grade language models on phones requires both small models and low-precision arithmetic, yet distillation and quantisation are usually applied one after the other. We train a 1.3B-parameter student with distillation losses computed through simulated 4-bit quantisation, so that the student learns to be robust to its own rounding error. Compared with distilling first and quantising afterwards, the joint procedure recovers 41% of the accuracy lost to 4-bit quantisation, reaches 94% of the full-precision student on a suite of assistant tasks, and runs 3.1× faster on a mid-range mobile processor.
Version 2 · 3 reviewer reports · 14 August 2026 · 10.xxxx/ctm.pb.2026.0008
Most agent benchmarks report one run per task, yet agents are stochastic. We analyse run-to-run variance on four public agent benchmarks and derive the number of repeated runs needed to detect a given difference between two systems at conventional power. For differences smaller than five points, single-run comparisons have less than 40% power on all four benchmarks. We provide a sequential testing procedure that stops early when the evidence is clear, and show that it halves the cost of repeated evaluation while keeping the false-positive rate below 5%.
Version 2 · 3 reviewer reports · 9 July 2026 · 10.xxxx/ctm.pb.2026.0007
Instruction-tuned models decline requests for many reasons, but little is known about how well their refusals track the actual risk of a request. We build a labelled set of 6,400 requests spanning benign, borderline and harmful categories and measure calibration of refusal decisions across eight open models. Refusal rates correlate only weakly with annotated risk for borderline requests, and over-refusal is concentrated in a handful of phrasing patterns rather than topics. A light-weight calibration step that uses the model's own confidence reduces over-refusal by 31% at no measurable cost to harmful-request refusal. We release the labelled set and the annotation guidelines.
Version 2 · 3 reviewer reports · 12 June 2026 · 10.xxxx/ctm.pb.2026.0004
Existing agent benchmarks measure whether a task is solved once, not whether it is solved reliably. CTM-Bench contains 1,200 tool-use tasks, each with five perturbed variants (rephrased instructions, reordered tool lists, injected API errors, latency spikes and truncated outputs) and reports pass^k, the probability that an agent succeeds on all k independent attempts. We document collection, annotation and licensing following datasheet guidelines, evaluate six open models, and find that pass^1 overstates reliability: the best model falls from 71% pass^1 to 38% pass^5. The benchmark, scoring code and the datasheet are released under CC BY 4.0.
Version 2 · 4 reviewer reports · 22 May 2026 · 10.xxxx/ctm.pb.2026.0003
As language-model agents move from demonstrations to deployments, the way they are evaluated determines which claims can be trusted. We review how agents are evaluated in the literature, organising practice along four axes: the task environment, the success criterion, the treatment of variance, and the reporting of cost. We identify recurring weaknesses: success is often judged by another model without validation, runs are rarely repeated, cost and latency are reported in under a third of studies, and environment versions are seldom pinned. We close with a checklist for evaluation reports and open problems, including measurement of safety-relevant side effects.
Version 2 · 3 reviewer reports · 6 March 2026 · 10.xxxx/ctm.pb.2026.0002
Parameter-efficient fine-tuning (PEFT) methods adapt large models by updating a small fraction of weights, but most evidence comes from a single model size. We fit scaling curves for low-rank adaptation across six base-model sizes spanning two orders of magnitude and four downstream tasks. The optimal adapter rank grows sub-linearly with model width, and the gap between PEFT and full fine-tuning shrinks as models grow, closing to within 0.6 points at the largest size. A simple rule that sets the rank proportional to the square root of the hidden dimension recovers 96% of the best rank found by grid search. We provide fitted coefficients, discuss threats to validity, and note that results on instruction-following tasks differ from those on classification.
