bestofn 1.0.0__tar.gz

bestofn-1.0.0/LICENSE

@@ -0,0 +1,216 @@
+
+                                 Apache License
+                           Version 2.0, January 2004
+                        http://www.apache.org/licenses/
+
+   TERMS AND CONDITIONS FOR USE, REPRODUCTION, AND DISTRIBUTION
+
+   1. Definitions.
+
+      "License" shall mean the terms and conditions for use, reproduction,
+      and distribution as defined by Sections 1 through 9 of this document.
+
+      "Licensor" shall mean the copyright owner or entity authorized by
+      the copyright owner that is granting the License.
+
+      "Legal Entity" shall mean the union of the acting entity and all
+      other entities that control, are controlled by, or are under common
+      control with that entity. For the purposes of this definition,
+      "control" means (i) the power, direct or indirect, to cause the
+      direction or management of such entity, whether by contract or
+      otherwise, or (ii) ownership of fifty percent (50%) or more of the
+      outstanding shares, or (iii) beneficial ownership of such entity.
+
+      "You" (or "Your") shall mean an individual or Legal Entity
+      exercising permissions granted by this License.
+
+      "Source" form shall mean the preferred form for making modifications,
+      including but not limited to software source code, documentation
+      source, and configuration files.
+
+      "Object" form shall mean any form resulting from mechanical
+      transformation or translation of a Source form, including but
+      not limited to compiled object code, generated documentation,
+      and conversions to other media types.
+
+      "Work" shall mean the work of authorship, whether in Source or
+      Object form, made available under the License, as indicated by a
+      copyright notice that is included in or attached to the work
+      (an example is provided in the Appendix below).
+
+      "Derivative Works" shall mean any work, whether in Source or Object
+      form, that is based on (or derived from) the Work and for which the
+      editorial revisions, annotations, elaborations, or other modifications
+      represent, as a whole, an original work of authorship. For the purposes
+      of this License, Derivative Works shall not include works that remain
+      separable from, or merely link (or bind by name) to the interfaces of,
+      the Work and Derivative Works thereof.
+
+      "Contribution" shall mean any work of authorship, including
+      the original version of the Work and any modifications or additions
+      to that Work or Derivative Works thereof, that is intentionally
+      submitted to Licensor for inclusion in the Work by the copyright owner
+      or by an individual or Legal Entity authorized to submit on behalf of
+      the copyright owner. For the purposes of this definition, "submitted"
+      means any form of electronic, verbal, or written communication sent
+      to the Licensor or its representatives, including but not limited to
+      communication on electronic mailing lists, source code control systems,
+      and issue tracking systems that are managed by, or on behalf of, the
+      Licensor for the purpose of discussing and improving the Work, but
+      excluding communication that is conspicuously marked or otherwise
+      designated in writing by the copyright owner as "Not a Contribution."
+
+      "Contributor" shall mean Licensor and any individual or Legal Entity
+      on behalf of whom a Contribution has been received by Licensor and
+      subsequently incorporated within the Work.
+
+   2. Grant of Copyright License. Subject to the terms and conditions of
+      this License, each Contributor hereby grants to You a perpetual,
+      worldwide, non-exclusive, no-charge, royalty-free, irrevocable
+      copyright license to reproduce, prepare Derivative Works of,
+      publicly display, publicly perform, sublicense, and distribute the
+      Work and such Derivative Works in Source or Object form.
+
+   3. Grant of Patent License. Subject to the terms and conditions of
+      this License, each Contributor hereby grants to You a perpetual,
+      worldwide, non-exclusive, no-charge, royalty-free, irrevocable
+      (except as stated in this section) patent license to make, have made,
+      use, offer to sell, sell, import, and otherwise transfer the Work,
+      where such license applies only to those patent claims licensable
+      by such Contributor that are necessarily infringed by their
+      Contribution(s) alone or by combination of their Contribution(s)
+      with the Work to which such Contribution(s) was submitted. If You
+      institute patent litigation against any entity (including a
+      cross-claim or counterclaim in a lawsuit) alleging that the Work
+      or a Contribution incorporated within the Work constitutes direct
+      or contributory patent infringement, then any patent licenses
+      granted to You under this License for that Work shall terminate
+      as of the date such litigation is filed.
+
+   4. Redistribution. You may reproduce and distribute copies of the
+      Work or Derivative Works thereof in any medium, with or without
+      modifications, and in Source or Object form, provided that You
+      meet the following conditions:
+
+      (a) You must give any other recipients of the Work or
+          Derivative Works a copy of this License; and
+
+      (b) You must cause any modified files to carry prominent notices
+          stating that You changed the files; and
+
+      (c) You must retain, in the Source form of any Derivative Works
+          that You distribute, all copyright, patent, trademark, and
+          attribution notices from the Source form of the Work,
+          excluding those notices that do not pertain to any part of
+          the Derivative Works; and
+
+      (d) If the Work includes a "NOTICE" text file as part of its
+          distribution, then any Derivative Works that You distribute must
+          include a readable copy of the attribution notices contained
+          within such NOTICE file, excluding those notices that do not
+          pertain to any part of the Derivative Works, in at least one
+          of the following places: within a NOTICE text file distributed
+          as part of the Derivative Works; within the Source form or
+          documentation, if provided along with the Derivative Works; or,
+          within a display generated by the Derivative Works, if and
+          wherever such third-party notices normally appear. The contents
+          of the NOTICE file are for informational purposes only and
+          do not modify the License. You may add Your own attribution
+          notices within Derivative Works that You distribute, alongside
+          or as an addendum to the NOTICE text from the Work, provided
+          that such additional attribution notices cannot be construed
+          as modifying the License.
+
+      You may add Your own copyright statement to Your modifications and
+      may provide additional or different license terms and conditions
+      for use, reproduction, or distribution of Your modifications, or
+      for any such Derivative Works as a whole, provided Your use,
+      reproduction, and distribution of the Work otherwise complies with
+      the conditions stated in this License.
+
+   5. Submission of Contributions. Unless You explicitly state otherwise,
+      any Contribution intentionally submitted for inclusion in the Work
+      by You to the Licensor shall be under the terms and conditions of
+      this License, without any additional terms or conditions.
+      Notwithstanding the above, nothing herein shall supersede or modify
+      the terms of any separate license agreement you may have executed
+      with Licensor regarding such Contributions.
+
+   6. Trademarks. This License does not grant permission to use the trade
+      names, trademarks, service marks, or product names of the Licensor,
+      except as required for reasonable and customary use in describing the
+      origin of the Work and reproducing the content of the NOTICE file.
+
+   7. Disclaimer of Warranty. Unless required by applicable law or
+      agreed to in writing, Licensor provides the Work (and each
+      Contributor provides its Contributions) on an "AS IS" BASIS,
+      WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or
+      implied, including, without limitation, any warranties or conditions
+      of TITLE, NON-INFRINGEMENT, MERCHANTABILITY, or FITNESS FOR A
+      PARTICULAR PURPOSE. You are solely responsible for determining the
+      appropriateness of using or redistributing the Work and assume any
+      risks associated with Your exercise of permissions under this License.
+
+   8. Limitation of Liability. In no event and under no legal theory,
+      whether in tort (including negligence), contract, or otherwise,
+      unless required by applicable law (such as deliberate and grossly
+      negligent acts) or agreed to in writing, shall any Contributor be
+      liable to You for damages, including any direct, indirect, special,
+      incidental, or consequential damages of any character arising as a
+      result of this License or out of the use or inability to use the
+      Work (including but not limited to damages for loss of goodwill,
+      work stoppage, computer failure or malfunction, or any and all
+      other commercial damages or losses), even if such Contributor
+      has been advised of the possibility of such damages.
+
+   9. Accepting Warranty or Additional Liability. While redistributing
+      the Work or Derivative Works thereof, You may choose to offer,
+      and charge a fee for, acceptance of support, warranty, indemnity,
+      or other liability obligations and/or rights consistent with this
+      License. However, in accepting such obligations, You may act only
+      on Your own behalf and on Your sole responsibility, not on behalf
+      of any other Contributor, and only if You agree to indemnify,
+      defend, and hold each Contributor harmless for any liability
+      incurred by, or claims asserted against, such Contributor by reason
+      of your accepting any such warranty or additional liability.
+
+   END OF TERMS AND CONDITIONS
+
+   APPENDIX: How to apply the Apache License to your work.
+
+      To apply the Apache License to your work, attach the following
+      boilerplate notice, with the fields enclosed by brackets "[]"
+      replaced with your own identifying information. (Don't include
+      the brackets!)  The text should be enclosed in the appropriate
+      comment syntax for the file format. We also recommend that a
+      file or class name and description of purpose be included on the
+      same "printed page" as the copyright notice for easier
+      identification within third-party archives.
+
+   Copyright [yyyy] [name of copyright owner]
+
+   Licensed under the Apache License, Version 2.0 (the "License");
+   you may not use this file except in compliance with the License.
+   You may obtain a copy of the License at
+
+       http://www.apache.org/licenses/LICENSE-2.0
+
+   Unless required by applicable law or agreed to in writing, software
+   distributed under the License is distributed on an "AS IS" BASIS,
+   WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+   See the License for the specific language governing permissions and
+   limitations under the License.
+
+   Copyright 2026 Alejandro Areces Rivera - Interlace AI
+
+   Licensed under the Apache License, Version 2.0 (the "License");
+   you may not use this file except in compliance with the License.
+   You may obtain a copy of the License at
+
+       http://www.apache.org/licenses/LICENSE-2.0
+
+   Unless required by applicable law or agreed to in writing, software
+   distributed under the License is distributed on an "AS IS" BASIS,
+   WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+   See the License for the specific language governing permissions and
+   limitations under the License.

bestofn-1.0.0/PKG-INFO

@@ -0,0 +1,351 @@
+Metadata-Version: 2.4
+Name: bestofn
+Version: 1.0.0
+Summary: Inference-time compute for language models: sample N reasoning trajectories and select the best.
+Author-email: Alejandro Areces Rivera <interlaceIA@gmail.com>
+License-Expression: Apache-2.0
+Project-URL: Homepage, https://huggingface.co/InterlaceAI/best-of-n
+Project-URL: Source, https://github.com/voidlinestudios12-jpg/Interlace-AI/tree/main/best-of-n
+Project-URL: Documentation, https://github.com/voidlinestudios12-jpg/Interlace-AI/blob/main/best-of-n/USAGE.md
+Project-URL: Paper, https://doi.org/10.5281/zenodo.21936833
+Project-URL: Issues, https://github.com/voidlinestudios12-jpg/Interlace-AI/issues
+Keywords: llm,best-of-n,inference-time-compute,test-time-compute,reasoning,verifier,self-consistency,majority-voting
+Classifier: Development Status :: 4 - Beta
+Classifier: Intended Audience :: Science/Research
+Classifier: Intended Audience :: Developers
+Classifier: Operating System :: OS Independent
+Classifier: Programming Language :: Python :: 3
+Classifier: Programming Language :: Python :: 3.8
+Classifier: Programming Language :: Python :: 3.9
+Classifier: Programming Language :: Python :: 3.10
+Classifier: Programming Language :: Python :: 3.11
+Classifier: Programming Language :: Python :: 3.12
+Classifier: Programming Language :: Python :: 3.13
+Classifier: Topic :: Scientific/Engineering :: Artificial Intelligence
+Classifier: Typing :: Typed
+Requires-Python: >=3.8
+Description-Content-Type: text/markdown
+License-File: LICENSE
+Provides-Extra: transformers
+Requires-Dist: torch; extra == "transformers"
+Requires-Dist: transformers; extra == "transformers"
+Provides-Extra: vllm
+Requires-Dist: vllm; extra == "vllm"
+Dynamic: license-file
+
+<div align="center">
+
+<img src="https://huggingface.co/InterlaceAI/best-of-n/resolve/main/figures/interlace-logo.png" width="96" alt="Interlace AI">
+
+**INTERLACE&nbsp;AI**
+
+# Best-of-N
+
+### Inference-time compute for any language model
+
+**Sample N reasoning trajectories from a frozen model and select the best one.**
+No weights are modified. All gains come from *how* the model is used.
+
+Works with **any causal LM** and **any N** — you configure both.
+
+[![DOI](https://zenodo.org/badge/DOI/10.5281/zenodo.21936833.svg)](https://doi.org/10.5281/zenodo.21936833)
+[![License](https://img.shields.io/badge/license-Apache%202.0-green)](https://github.com/voidlinestudios12-jpg/Interlace-AI/blob/main/best-of-n/../LICENSE)
+[![Tests](https://img.shields.io/badge/tests-45%20passing-brightgreen)](https://github.com/voidlinestudios12-jpg/Interlace-AI/blob/main/best-of-n/tests/test_selectors.py)
+
+</div>
+
+---
+
+## What it does
+
+A language model does not produce *an* answer. It produces a **distribution over
+answers**, and generating text draws one sample from it. Ask the same question
+twice at non-zero temperature and you can get two different results.
+
+That variability is usually treated as a nuisance. Best-of-N treats it as a
+resource.
+
+Suppose a model solves a given problem correctly 30% of the time. Ask once and
+you are right 30% of the time. Ask 32 times and the probability that **at least
+one** attempt is correct is `1 − 0.7³² ≈ 99.99%`. The knowledge is there; a
+single sample just fails to retrieve it reliably.
+
+So the problem splits in two:
+
+| | |
+|---|---|
+| **Coverage** | Did any of the N attempts get it right? `1 − (1−p)^N` |
+| **Selection** | Did we manage to pick that one out of the N? |
+
+This library implements both halves: sampling N trajectories, and four different
+ways of choosing between them.
+
+![A 1.5B model against the field](https://huggingface.co/InterlaceAI/best-of-n/resolve/main/figures/04_vs_other_models.png)
+
+Against published results from other laboratories. Ours is majority vote at
+N=128 — the figure a system can actually return; the others are their published
+pass@1. Coverage is shown separately and labelled as a ceiling.
+
+![AIME results](https://huggingface.co/InterlaceAI/best-of-n/resolve/main/figures/01_aime_base_vs_bestofn.png)
+
+Measured on `DeepSeek-R1-Distill-Qwen-1.5B`, **AIME 2024**, weights untouched:
+
+| N | Majority vote | Coverage (pass@N) |
+|---:|---:|---:|
+| 1 | 23.3% | 23.3% |
+| 8 | 40.0% | 60.0% |
+| 32 | 50.0% | 73.3% |
+| **128** | **53.3%** | **83.3%** |
+
+![Across benchmarks](https://huggingface.co/InterlaceAI/best-of-n/resolve/main/figures/02_three_benchmarks.png)
+
+And across domains:
+
+| Benchmark | N | Single sample | Majority vote | Δ |
+|---|---:|---:|---:|---:|
+| AIME 2024 | 128 | 23.3% | **53.3%** | +30.0 |
+| GPQA-Diamond | 32 | 33.8% | **43.4%** | +9.6 |
+| GSM8K | 4 | 87.2% | **92.8%** | +5.6 |
+
+All three are majority vote — the selector you would actually deploy. AIME
+coverage at N=128 is higher (83.3%) but is a ceiling, not a returnable result.
+
+### How a call works
+
+1. **Generate.** N reasoning trajectories are sampled in parallel from the frozen
+   model at `temperature > 0`, so each explores a different path.
+2. **Extract.** The final answer is pulled out of each trajectory (by default the
+   last `oxed{...}`) and normalised, so `"204"`, `"204.0"` and `" 204 "` count
+   as the same answer.
+3. **Select.** A selector decides which answer to return — by frequency, by the
+   model's own confidence, or by an external verifier's score.
+4. **Return.** You get the answer plus every sample, so you can inspect them or
+   apply a different selector for free.
+
+---
+
+## Quickstart
+
+```python
+from bestofn import BestOfN
+
+engine = BestOfN("deepseek-ai/DeepSeek-R1-Distill-Qwen-1.5B", n=32)
+
+result = engine.solve("What is the remainder when 7^100 is divided by 13?")
+print(result.answer)
+print(result.agreement)     # how much the 32 samples agreed
+```
+
+Any model, any N:
+
+```python
+BestOfN("Qwen/Qwen2.5-Math-7B-Instruct", n=64)
+BestOfN("meta-llama/Llama-3.1-8B-Instruct", n=16)
+BestOfN("/path/to/your/local/model", n=128)
+```
+
+Install:
+
+```bash
+pip install torch transformers
+pip install vllm     # recommended; required for large N
+```
+
+---
+
+## The selection gap
+
+Look at the table above again. At N=128 the model **reaches** the right answer
+83.3% of the time, but majority voting only **returns** it 53.3% of the time.
+
+![The selection gap](https://huggingface.co/InterlaceAI/best-of-n/resolve/main/figures/03_selection_gap.png)
+
+That 30-point difference is the **selection gap** — the correct answer was
+generated and then discarded, because it was in the minority.
+
+Majority voting cannot fix this: it is structurally incapable of choosing an
+answer most trajectories disagree with. A learned verifier can.
+
+| Selector (N=32, 90 AIME problems) | Accuracy |
+|---|---:|
+| Self-certainty | 18.9% |
+| Majority vote | 35.6% |
+| Verifier — argmax trajectory | 43.3% |
+| **Verifier — confidence-weighted vote** | **52.2%** |
+
+```python
+def my_verifier(problem: str, trajectory: str) -> float:
+    return probability_it_is_correct        # 0.0 – 1.0
+
+engine = BestOfN(model, n=32, verifier=my_verifier)
+engine.solve(problem, method="verifier")     # +16.6 points over majority
+```
+
+---
+
+## Selectors
+
+| Method | Needs | Recovers a minority-correct answer? |
+|---|---|:---:|
+| `majority` | nothing | No |
+| `self_certainty` | log-probs (automatic) | Rarely |
+| `verifier` | a verifier callable | **Yes** |
+| `verifier_argmax` | a verifier callable | Yes, but noisier |
+| `oracle` | the gold answer | Diagnostic ceiling only |
+
+Generation is the expensive part — reusing the samples with another selector is free:
+
+```python
+r = engine.solve(problem, n=32)
+r.answer                          # majority
+r.select_with("self_certainty")
+r.covered(gold)                   # was the answer reachable at all? (pass@N)
+```
+
+---
+
+## Configuration
+
+```python
+BestOfN(
+    model="deepseek-ai/DeepSeek-R1-Distill-Qwen-1.5B",
+    n=32,                  # your compute budget
+    temperature=0.6,       # must be > 0
+    top_p=0.95,
+    max_tokens=8192,
+    extractor="boxed",     # boxed | number | letter | regex | your callable
+    backend="auto",        # auto | vllm | transformers
+    verifier=None,
+)
+```
+
+**Full documentation: [USAGE.md](https://github.com/voidlinestudios12-jpg/Interlace-AI/blob/main/best-of-n/USAGE.md)** — how to choose N, how to plug in a
+verifier, how to measure the selection gap on your own task, and the failure
+modes worth knowing about.
+
+---
+
+## Verification
+
+The published numbers are not asserted, they are replayed. This repository ships
+the trajectories that produced them and a script that feeds them back through
+the library:
+
+```bash
+python tests/verify_against_measured.py
+```
+
+```
+Samples per problem (N)          : 32
+  majority selection reproduced  : 30/30  (100.0%)
+  majority vote, this library    : 63.3%   [recorded run: 63.3%]
+  coverage pass@32                 : 76.7%
+  gain from Best-of-N            : +31.7 points
+PASS - reproduces the published measurements exactly
+```
+
+> **Why 63.3% here and 53.3% in the tables above?** They are two independent runs
+> of the same benchmark. AIME has only 30 problems, so a single run carries a
+> standard error of roughly 9 points; our two runs differ by 13 points at N=32
+> (63.3% vs 50.0%). The tables quote the **more conservative** run. This script
+> replays the other one, which is the set of trajectories shipped in `tests/`.
+> Both are real measurements — the spread is what a 30-problem benchmark looks
+> like, and it is worth knowing before comparing any single AIME figure against
+> another lab'"'"'s.
+
+Plus 45 unit tests covering extraction, normalisation, every selector, the
+minority-rescue mechanism and error handling — no GPU required:
+
+```bash
+python tests/test_selectors.py     # 45 passed, 0 failed
+```
+
+---
+
+## When this helps, and when it does not
+
+**Good fit**
+
+- Problems with a single comparable final answer: maths, multiple choice, short
+  factual questions, unit-testable code.
+- A model that is *sometimes* right. The gain is largest when per-sample accuracy
+  sits in the middle — around 20–60%.
+- Accuracy matters more than latency, and you can afford N generations.
+
+**Poor fit**
+
+- Open-ended generation — essays, summaries, chat. There is no well-defined vote
+  over free text.
+- Problems the model never solves. If `p = 0`, then `1 − (1−p)^N = 0` for every
+  N: sampling more changes nothing.
+- Problems the model always solves. Nothing to recover; you are paying N times
+  for the same answer.
+- Hard latency budgets. Cost scales linearly with N.
+
+Not sure which case you are in? Measure it:
+
+```python
+results = engine.solve_batch(problems, n=32)
+selected  = sum(r.answer == g for r, g in zip(results, golds)) / len(golds)
+reachable = sum(r.covered(g)  for r, g in zip(results, golds)) / len(golds)
+print(f"selected {selected:.1%} · reachable {reachable:.1%} · gap {reachable-selected:.1%}")
+```
+
+A large gap means invest in a verifier. Low coverage means you need a better
+base model, not more samples.
+
+---
+
+## Limitations
+
+- **Coverage is a hard ceiling.** If the model never generates the correct
+  answer, no selector recovers it: `1 − (1−p)^N` is 0 for every N when p = 0.
+- **Cost is linear in N.** N=128 means 128 generations. Use `vllm`.
+- **Needs an extractable answer.** Built for tasks with a comparable final
+  answer. Open-ended generation has no well-defined vote.
+- **Verifier errors amplify.** The selector is bounded by its verifier.
+
+---
+
+## Prior work
+
+Best-of-N sampling and verifier-based reranking are established techniques. This is an open
+implementation with published measurements, not a new method:
+
+- Cobbe et al., *Training Verifiers to Solve Math Word Problems*, 2021 — verifier reranking of N samples
+- Wang et al., *Self-Consistency Improves Chain of Thought Reasoning*, 2022 — majority voting over samples
+- Lightman et al., *Let'''s Verify Step by Step*, 2023 — process vs outcome supervision
+- Snell et al., *Scaling LLM Test-Time Compute Optimally*, 2024 — compute vs parameters
+- Brown et al., *Large Language Monkeys*, 2024 — coverage scaling with repeated sampling
+
+Full discussion and references in [the technical report](https://doi.org/10.5281/zenodo.21936833).
+
+---
+
+## Citation
+
+```bibtex
+@techreport{arecesrivera2026interlace,
+  title  = {Modern Architecture On Advanced LLM: Best-of-N Sampling,
+            Learned Verification and Tree Search as a Substitute for Parameter Scale},
+  author = {Areces Rivera, Alejandro},
+  year   = {2026},
+  number = {TR-2026-01},
+  institution = {Interlace AI},
+  doi    = {10.5281/zenodo.21936833}
+}
+```
+
+Technical report: [doi.org/10.5281/zenodo.21936833](https://doi.org/10.5281/zenodo.21936833)
+Code and raw outputs: [github.com/.../Interlace-AI/best-of-n](https://github.com/voidlinestudios12-jpg/Interlace-AI/tree/main/best-of-n)
+
+---
+
+## License
+
+**Apache License 2.0** — free to use, modify and redistribute, including
+commercially. Use it with any model, at any N, in any project.
+
+Copyright 2026 Alejandro Areces Rivera — Interlace AI
+
+Questions and collaboration: `interlaceIA@gmail.com`