`包裹，长文用`
`/`
`划分小节。 3. 列表用`
- `，引用用`
`，代码用`
`，重要句子用``。
4. 保留原文所有信息，不得增删改内容。
5. 不要输出``/``/``等文档级标签。
* 输入文本：一段关于药物开发、强化学习、ReMedI方法的学术摘要/新闻稿。
2. 分析输入文本结构：
* "In drug development, lead optimization must improve competing properties, such as increasing permeability and reducing toxicity, while limiting scaffold changes that could compromise the lead's potency. RL-based post-training, such as C-MORAL, addresses this problem by fine-tuning molecular language models to directly optimize molecular leads for objectives of interest, which relies on expensive task-specific datasets and faces challenges in achieving multi-objective optimization (MOO). We argue a pre-trained model may already contain useful molecular edits that zero-shot generation cannot efficiently reach. Motivated by the limited search capabilities of zero-shot generation, we develop an efficient Metropolis-Hastings search strategy to better explore the proposal distribution." -> 背景与问题（可以作为引言或第一段，或者加个小标题）。
* "We propose ReMedI, which uses the molecular language model as a base proposer for inference-time alignment. Given a lead and a set of multi-objective tasks, the model proposes edits conditioned on the current molecule, while Metropolis--Hastings refinement guides sequential edits under an objective for property improvement, preservation, and lead similarity." -> 方法介绍（ReMedI）。
* "On both C-MuMOInstruct and SMDD benchmarks, refinement improves frozen base models across three backbones, reaching significantly higher multi-objective optimization success rates compared to baselines. These search traces can then distilled into the base model which improves zero-shot generation and subsequent search using approximately 7.8 times less training molecular leads than that of C-MORAL. ReMedI achieves approximately 2.8 times the strict-success coverage of RL baseline C-MORAL, including on out-of-distribution tasks, with matched lead-relative similarity." -> 实验结果与结论。
3. 设计HTML结构：
* 由于是一篇短文，可以加一个主标题（比如 `` 或保持段落），规则说“长文用`
`/`
`划分小节”，这里可以稍微分几个小节以体现排版能力。
* 小节1：背景与挑战 (`
背景与挑战
` 或者英文？原文是英文，最好保持原文语言，或者用英文标题，这里用英文标题更自然，或者直接不加标题只分段？规则说“长文用/
划分小节”，我加几个英文小标题来切分逻辑。)
* `
Background and Challenges
` -> 第一段。把重要句子加粗，比如 "We argue a pre-trained model may already contain useful molecular edits that zero-shot generation cannot efficiently reach."
* 小节2：The ReMedI Approach (`The ReMedI Approach
`) -> 第二段。
* 小节3：Experimental Results (`Experimental Results
`) -> 第三段。
* *等等，规则4说“保留原文所有信息，不得增删改内容”。如果我加了小标题（如 "Background and Challenges"），这算不算“增删改内容”？*
* 为了严格遵守“不得增删改内容”，我不应该添加原文没有的小标题文本。如果非要用 ``/`
`，我只能把原文的某些句子提升为标题，但这会改变原文的句法结构或者看起来像标题。
* 更好的做法是：直接使用 `
` 包裹所有自然段落。如果原文没有明确的章节标题，就不强行造标题。或者，我可以把第一句话或者某些关键句用 `` 标注。
* 再仔细看规则2：“每个自然段落用
包裹，长文用
`，代码用``，重要句子用``。 4. 保留原文所有信息，不得增删改内容。 5. 不要输出``/``/``等文档级标签。 * 输入文本：一段关于药物开发、强化学习、ReMedI方法的学术摘要/新闻稿。 2. 分析输入文本结构： * "In drug development, lead optimization must improve competing properties, such as increasing permeability and reducing toxicity, while limiting scaffold changes that could compromise the lead's potency. RL-based post-training, such as C-MORAL, addresses this problem by fine-tuning molecular language models to directly optimize molecular leads for objectives of interest, which relies on expensive task-specific datasets and faces challenges in achieving multi-objective optimization (MOO). We argue a pre-trained model may already contain useful molecular edits that zero-shot generation cannot efficiently reach. Motivated by the limited search capabilities of zero-shot generation, we develop an efficient Metropolis-Hastings search strategy to better explore the proposal distribution." -> 背景与问题（可以作为引言或第一段，或者加个小标题）。 * "We propose ReMedI, which uses the molecular language model as a base proposer for inference-time alignment. Given a lead and a set of multi-objective tasks, the model proposes edits conditioned on the current molecule, while Metropolis--Hastings refinement guides sequential edits under an objective for property improvement, preservation, and lead similarity." -> 方法介绍（ReMedI）。 * "On both C-MuMOInstruct and SMDD benchmarks, refinement improves frozen base models across three backbones, reaching significantly higher multi-objective optimization success rates compared to baselines. These search traces can then distilled into the base model which improves zero-shot generation and subsequent search using approximately 7.8 times less training molecular leads than that of C-MORAL. ReMedI achieves approximately 2.8 times the strict-success coverage of RL baseline C-MORAL, including on out-of-distribution tasks, with matched lead-relative similarity." -> 实验结果与结论。 3. 设计HTML结构： * 由于是一篇短文，可以加一个主标题（比如 `` 或保持段落），规则说“长文用`
`/`
`划分小节”，这里可以稍微分几个小节以体现排版能力。 * 小节1：背景与挑战 (`
背景与挑战
` 或者英文？原文是英文，最好保持原文语言，或者用英文标题，这里用英文标题更自然，或者直接不加标题只分段？规则说“长文用/
划分小节”，我加几个英文小标题来切分逻辑。) * `
Background and Challenges
` -> 第一段。把重要句子加粗，比如 "We argue a pre-trained model may already contain useful molecular edits that zero-shot generation cannot efficiently reach." * 小节2：The ReMedI Approach (`The ReMedI Approach
`) -> 第二段。 * 小节3：Experimental Results (`Experimental Results
`) -> 第三段。 * *等等，规则4说“保留原文所有信息，不得增删改内容”。如果我加了小标题（如 "Background and Challenges"），这算不算“增删改内容”？* * 为了严格遵守“不得增删改内容”，我不应该添加原文没有的小标题文本。如果非要用 ``/`
`，我只能把原文的某些句子提升为标题，但这会改变原文的句法结构或者看起来像标题。 * 更好的做法是：直接使用 `
` 包裹所有自然段落。如果原文没有明确的章节标题，就不强行造标题。或者，我可以把第一句话或者某些关键句用 `` 标注。 * 再仔细看规则2：“每个自然段落用
包裹，长文用
/划分小节”。如果不用加标题，我就直接用 `
`。如果必须用，我可以把原文的逻辑断句用 `
` 包起来吗？不行，会改变原文。所以直接全部用 `
` 包裹是最安全的，或者把某些长段落拆成多个 `
`。 * 让我们把这段文本分成3个 `
`。 * 第一段：In drug development... proposal distribution. * 第二段：We propose ReMedI... lead similarity. * 第三段：On both C-MuMOInstruct... lead-relative similarity. * 在段落内部，可以使用 `` 来强调重要句子，符合规则3。 * 例如： * 第一段中的 "We argue a pre-trained model may already contain useful molecular edits that zero-shot generation cannot efficiently reach." 加粗。 * 第二段中的 "We propose ReMedI, which uses the molecular language model as a base proposer for inference-time alignment." 加粗。 * 第三段中的 "ReMedI achieves approximately 2.8 times the strict-success coverage of RL baseline C-MORAL, including on out-of-distribution tasks, with matched lead-relative similarity." 加粗。 4. 细化HTML标签使用： * 规则3提到：列表用`
- `，引用用`
`，代码用`
`，重要句子可用``。
* 这段文本里没有明显的列表、引用或代码。所以只用 `` 和 ``。
* 是否需要用 `
`/``？如果不用会不会违反规则？“长文用
/
划分小节”，这段文本不长（大约150词），用 `
` 即可。如果非要体现排版能力，我可以把段落首句或者某些关键概念用 `
` 包起来？不，这会破坏句子结构。所以只用 `
` 和 ``。 * 让我们检查一下原文是否有可以拆分为列表的部分。 * "objectives for property improvement, preservation, and lead similarity." -> 可以变成列表吗？规则4“不得增删改内容”，把句子里的并列词拆成列表可能会被视为修改了原文的文本结构（虽然语义没变，但为了安全起见，保持原句）。 * 那就老老实实用 `
` 和 ``。 5. 草拟HTML： ```html
In drug development, lead optimization must improve competing properties, such as increasing permeability and reducing toxicity, while limiting scaffold changes that could compromise the lead's potency. RL-based post-training, such as C-MORAL, addresses this problem by fine-tuning molecular language models to directly optimize molecular leads for objectives of interest, which relies on expensive task-specific datasets and faces challenges in achieving multi-objective optimization (MOO). We argue a pre-trained model may already contain useful molecular edits that zero-shot generation cannot efficiently reach. Motivated by the limited search capabilities of zero-shot generation, we develop an efficient Metropolis-Hastings search strategy to better explore the proposal distribution.
We propose ReMedI, which uses the molecular language model as a base proposer for inference-time alignment. Given a lead and a set of multi-objective tasks, the model proposes edits conditioned on the current molecule, while Metropolis--Hastings refinement guides sequential edits under an objective for property improvement, preservation, and lead similarity.
On both C-MuMOInstruct and SMDD benchmarks, refinement improves frozen base models across three backbones, reaching significantly higher multi-objective optimization success rates compared to baselines. These search traces
`，代码用``，重要句子可用``。 * 这段文本里没有明显的列表、引用或代码。所以只用 `` 和 ``。 * 是否需要用 `
`/`
`？如果不用会不会违反规则？“长文用
/
划分小节”，这段文本不长（大约150词），用 `
` 即可。如果非要体现排版能力，我可以把段落首句或者某些关键概念用 `
` 包起来？不，这会破坏句子结构。所以只用 `
` 和 ``。 * 让我们检查一下原文是否有可以拆分为列表的部分。 * "objectives for property improvement, preservation, and lead similarity." -> 可以变成列表吗？规则4“不得增删改内容”，把句子里的并列词拆成列表可能会被视为修改了原文的文本结构（虽然语义没变，但为了安全起见，保持原句）。 * 那就老老实实用 `
` 和 ``。 5. 草拟HTML： ```html
In drug development, lead optimization must improve competing properties, such as increasing permeability and reducing toxicity, while limiting scaffold changes that could compromise the lead's potency. RL-based post-training, such as C-MORAL, addresses this problem by fine-tuning molecular language models to directly optimize molecular leads for objectives of interest, which relies on expensive task-specific datasets and faces challenges in achieving multi-objective optimization (MOO). We argue a pre-trained model may already contain useful molecular edits that zero-shot generation cannot efficiently reach. Motivated by the limited search capabilities of zero-shot generation, we develop an efficient Metropolis-Hastings search strategy to better explore the proposal distribution.
We propose ReMedI, which uses the molecular language model as a base proposer for inference-time alignment. Given a lead and a set of multi-objective tasks, the model proposes edits conditioned on the current molecule, while Metropolis--Hastings refinement guides sequential edits under an objective for property improvement, preservation, and lead similarity.
On both C-MuMOInstruct and SMDD benchmarks, refinement improves frozen base models across three backbones, reaching significantly higher multi-objective optimization success rates compared to baselines. These search traces