Implementation:Testtimescaling Testtimescaling github io HTML Table Row Append

Type	Pattern Doc (manual edit process)
Source	`README.md:L75-106` (Paper Tables section)
Domains	Documentation, Data_Management
Last Updated	2026-02-14

Overview

A manual editing process for appending a new paper row to the HTML comparison table in README.md, following the established column format with arXiv badges and taxonomy values.

Description

This pattern documents the exact format and process for adding a new row to the comparison table in the repository's README.md. The table uses GitHub-flavored Markdown table syntax with inline HTML for formatting paper titles and arXiv badges.

The contributor must:

Open README.md and navigate to the paper comparison table (approximately lines 75-106).
Identify the last row in the table.
Append a new row following the exact format template.
Ensure all taxonomy values from the classification step are correctly placed in the corresponding columns.

The table header structure (already present in README.md) defines the column layout that every row must follow:

| <div style="width:300px">Method(PapersTitles)</div> | What | How → |        |        |        |        |        | Where | How Well |
|--------|------|-------|--------|--------|--------|--------|--------|-------|-------|
|        |      | SFT   | RL     | STI | SEA | VER | AGG |        |        |

Usage

Use this process after completing taxonomy classification (Step 2). Have the following information ready before editing:

Paper title (exact title from the paper)
arXiv ID (format: XXXX.XXXXX)
All nine taxonomy classification values (What, SFT, RL, STI, SEA, VER, AGG, Where, How Well)

Code Reference

Source Location

README.md:L75-106 in the repository root. The table is located in the section following the taxonomy description.

Interface Specification

The row format template is:

|<i><b>Paper Title</b></i>, <a href="https://arxiv.org/abs/XXXX.XXXXX" target="_blank"><img src="https://img.shields.io/badge/arXiv-XXXX.XXXXX-red" alt="arXiv Badge"></a></li>|What_Value|SFT_Value|RL_Value|STI_Value|SEA_Value|VER_Value|AGG_Value|Where_Value|How_Well_Value|

Where:

Paper Title is the full paper title, wrapped in <i><b>...</b></i>
XXXX.XXXXX is the arXiv ID, appearing in both the href URL and the badge src URL
Each *_Value is the taxonomy classification value or ✗ for not applicable
Values with special characters should use HTML entities where needed

Field value conventions:

Field	Example Values	Not Applicable
What	Parallel, Sequential, Hybrid, Internal	(always has a value)
SFT	Distillation, Instruction Tuning	✗
RL	GRPO, DPO, PPO, REINFORCE	✗
STI	CoT, Self-Refine, Budget Forcing	✗
SEA	MCTS, Beam, Best-First	✗
VER	PRM, ORM, Self-Evaluate	✗
AGG	Best-of-N, Majority Vote, Fusion	✗
Where	Math, Code, Sci, Game, Open-Ended	(always has a value)
How Well	Pass@1, Accuracy, Win Rate	(always has a value)

Import

No imports required. This is a manual file edit performed in any text editor or through the GitHub web interface.

I/O Contract

Inputs

Parameter	Type	Required	Description
paper_title	String	Yes	The full title of the paper
arxiv_url	String	Yes	The full arXiv URL (e.g., `https://arxiv.org/abs/XXXX.XXXXX`)
arxiv_id	String	Yes	The arXiv identifier (e.g., `XXXX.XXXXX`)
classification	Object	Yes	All taxonomy values from Step 2 (what, sft, rl, sti, sea, ver, agg, where, how_well)

Outputs

Output	Type	Description
updated_readme	File	The `README.md` file with the new row appended to the comparison table
rendered_row	HTML	The new table row visible in the GitHub-rendered README

Usage Examples

Example 1: Adding a search-based paper

|<i><b>Scaling LLM Test-Time Compute Optimally can be More Effective than Scaling Model Parameters</b></i>, <a href="https://arxiv.org/abs/2408.03314" target="_blank"><img src="https://img.shields.io/badge/arXiv-2408.03314-red" alt="arXiv Badge"></a></li>|Sequential|✗|✗|✗|Beam|PRM|Best-of-N|Math|Pass@1, Accuracy|

Example 2: Adding an RL-trained reasoning model

|<i><b>DeepSeek-R1: Incentivizing Reasoning Capability in LLMs via Reinforcement Learning</b></i>, <a href="https://arxiv.org/abs/2501.12948" target="_blank"><img src="https://img.shields.io/badge/arXiv-2501.12948-red" alt="arXiv Badge"></a></li>|Sequential|Cold Start SFT|GRPO|Long CoT|✗|✗|✗|Math, Code, Sci|Pass@1, Accuracy|

Example 3: Adding a parallel aggregation paper

|<i><b>Self-Consistency Improves Chain of Thought Reasoning in Language Models</b></i>, <a href="https://arxiv.org/abs/2203.11171" target="_blank"><img src="https://img.shields.io/badge/arXiv-2203.11171-red" alt="arXiv Badge"></a></li>|Parallel|✗|✗|CoT|✗|✗|Majority Vote|Math, Code|Accuracy|

Related Pages

Page Connections

Double-click a node to navigate. Hold to expand connections.

Principle

Implementation

Heuristic

Environment