Bibliographic record
Abstract
Contributors to this version: Abel Aoun (@bzah), Pascal Bourgault (@aulemahal), David Huard (@huard), Juliette Lavoie (@juliettelavoie), Travis Logan (@tlogan2000), Trevor James Smith (@Zeitsperre). New indicators <code>thawing_degree_days</code> indicator returns degree-days above a default of <code>thresh="0 degC"</code>. (PR/895, GH/887). <code>freezing_degree_days</code> indicator returns degree-days below a default of <code>thresh="0 degC"</code>. (PR/895, GH/887). Several frost-free season calculations are now available as both indices and indicators. (PR/895, GH/887): <code>frost_free_season_start</code> <code>frost_free_season_end</code> <code>frost_free_season_length</code> <code>growing_season_start</code> is now offered as an indice and as an indicator to complement other growing season-based indicators (threshold calculation with <code>op=">="</code>). (PR/895, GH/887). New features and enhancements Improve cell_methods checking to search the wanted method within the whole string. (PR/866, GH/863). New <code>align_on='random</code> option for <code>xclim.core.calendar.convert_calendar</code>, for conversions involving '360_day' calendars. (PR/875, GH/841). <code>dry_spell_frequency</code> now has a parameter <code>op: {"sum", "max"}</code> to choose if the threshold is compared against the accumulated or maximal precipitation, over the given window. (PR/879). <code>maximum_consecutive_frost_free_days</code> is now checking that the minimum temperature is above or equal to the threshold ( instead of only above). (PR/883, GH/881). The ANUCLIM virtual module has been updated to accept weekly and monthly inputs and with improved metadata. (PR/885, GH/538) The <code>sdba.loess</code> algorithm has been optimized to run faster in all cases, with an even faster special case (<code>equal_spacing=True</code>) when the x coordinate is equally spaced. When activated, this special case might return results different from without, up to around 0.1%. (PR/865). Add support for group's window and additional dimensions in <code>LoessDetrend</code>. Add new <code>RollingMeanDetrend</code> object. (PR/865). Missing value algorithms now try to infer the source timestep of the input data when it is not given. (PR/885). On indices, <code>bootstrap</code> parameter documentation has been updated to explain when and why it should be used. (PR/893, GH/846). Breaking changes Major changes in the YAML schema for virtual submodules, now closer to how indicators are declared dynamically, see the doc for details. (PR/849, GH/848). Removed <code>xclim.generic.daily_downsampler</code>, as it served no purpose now that xarray's resampling works with cftime (PR/888, GH/889). Refactor of <code>xclim.core.calendar.parse_offset</code>, output types were changed to useful ones (PR/885). Major changes on how parameters are passed to indicators. (PR/873): Their signature is now consistent : input variables (DataArrays, optional or not) are positional or keyword arguments and all other parameters are keyword only. (GH/855, GH/857) Some indicators have modified signatures because we now rename variables when wrapping generic indices. This is the case for the whole cf module, for example. <code>Indicator.parameters</code> is now a property generated from <code>Indicator._all_parameters</code>, as the latter includes the injected parameters. The keys of the former are instances of new <code>xclim.core.indicator.Parameter</code>, and not dictionaries as before. New <code>Indicator.injected_parameters</code> to see which compute function arguments will be injected at call time. See the pull request (PR/873) for all information. The call signature for <code>huglin_index</code> has been modified to reflect the correct variables used in its formula (<code>tasmin</code> -> <code>tas</code>; <code>thresh_tasmin</code> -> <code>thresh</code>). (PR/903, GH/902). Internal changes Pull Request contributions now require hyperlinks to the issue and pull request pages on GitHub listed alongside changess in HISTORY.rst. (PR/860,GH/854). Updated the contribution guidelines to better give credit to contributors and more easily track changes. (PR/869, GH/868). Enabled coveralls code coverage reporting for GitHub CI. (PR/870). Added automated TestPyPI and PyPI-publishing workflows for GitHub CI. (PR/872). Changes on how indicators are constructed. (PR/873). Added missing algorithms tests for conversion from hourly to daily. (PR/888). Updated pre-commit hooks to use black v21.10.b0. (PR/896). Moved <code>stack_variables</code>, <code>unstack_variables</code>, <code>construct_moving_yearly_window</code> and <code>unpack_moving_yearly_window</code> from <code>xclim.sdba.base</code> to <code>xclim.sdba.processing</code>. They still are imported in <code>xclim.sdba</code> as before. (PR/892). Many improvements to the documentation. (PR/892, GH/880). Added regex replacement handling in setup.py to facilitate publishing contributor/contribution links on PyPI. (PR/906). Bug fixes Fix a bug in bootstrapping where computation would fail when the dataset time coordinate is encoded using <code>cftime.datetime</code>. (PR/859). Fix a bug in <code>build_indicator_module_from_yaml</code> where bases classes (Daily, Hourly, etc) were not usable with the <code>base</code> field. (PR/885). <code>percentile_doy</code> alpha and beta parameters are now properly transmitted to bootstrap calls of this function. (PR/893, GH/846). When called with a 1D da and ND index, <code>xclim.indices.run_length.lazy_indexing</code> now drops the auxiliary coordinate corresponding to da's index. This fixes a bug with ND data in <code>xclim.indices.run_length.season</code>. (PR/900). Fix name of heating degree days in French (<code>"chauffe"</code> -> "<code>chauffage</code>"). (PR/895). Corrected several French indicator translation description strings (bad usages of <code>"."</code> in <code>description</code> and <code>long_name</code> fields). (PR/895). Fixed an error with the formula for <code>huglin_index</code> where <code>tasmin</code> was being used in the calculation instead of <code>tas</code>. (PR/903, GH/902).
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
How this classification was reachedexpand
Full frame distilled prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. Learned from the 10,348 direct Codex labels and 10,348 direct Gemma labels. Candidate is the union of thresholded teacher heads; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels or direct frontier model labels.
Codex and Gemma teacher scores by category
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.000 | 0.000 |
| Meta-epidemiology (narrow) | 0.000 | 0.000 |
| Meta-epidemiology (broad) | 0.000 | 0.000 |
| Bibliometrics | 0.000 | 0.001 |
| Science and technology studies | 0.002 | 0.000 |
| Scholarly communication | 0.001 | 0.000 |
| Open science | 0.001 | 0.002 |
| Research integrity | 0.000 | 0.001 |
| Insufficient payload (model declined to judge) | 0.189 | 0.018 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from itClassification
machine, unvalidatedMachine predicted; both teacher heads agree on what is shown here.
How this classification was reached, model by model and score by score, is at the end of the page under "How this classification was reached".