Dawei Li, Zhao Yang, Yuxuan Jiang, Runchao Li, Shubhashis Dipta
11 of 11 samples (functions and classes) lifted from this paper's linked repositories executed on a synthesized input in Syntology's jail (0 of them constructed an object rather than computing a result; 4 honoured the contract we drafted, 0 violated it and 0 have no contract verdict; 7 failures of our instrument; 0 no recorded run), measured 2026-11-22; bodies inline only under a permissive licence as GitHub detects it. Not a reproduction of the paper's results.
"Ran" means the function executed on a synthesized input and returned a value. It is not a reproduction of the paper's results. This record, dated, as JSON: /api/ran/2605.09253.json · badge: /api/ran/2605.09253.svg
Each repository's role is a label: a classifier's output recorded on the link, or, where no role is recorded, what the Papers with Code archive marked. The line under it names its basis. Neither says who wrote the code.
| Repository | Role | Ran |
|---|---|---|
| YuxuanJiang1/Rock-Token | canonical (classifier label) classifier basis: census |
11 ran · 0 no recorded run |
| Function | Status | Where it lives |
|---|---|---|
| binned_envelope | Ran ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT) |
YuxuanJiang1/Rock-Token/rock_detection/funnel_plot.py pointer only (licence: NONE) · get_code("e9383e9db51e9cd4") How it ranThe body of this sample is withheld: GitHub detected no licence in this paper's repository, YuxuanJiang1/Rock-Token when it was checked. The setup block that called it is withheld with it, because it is written against that body. source_path and origin say where the code lives, to read under the repository's own terms. A run is recorded. |
| cos_to | Ran ran; the contract check could not run (RAISES, DEP_MISSING or TIMEOUT) |
YuxuanJiang1/Rock-Token/rock_detection/analyze_gradient_alignment.py pointer only (licence: NONE) · get_code("beb0c035178abca6") How it ranThe body of this sample is withheld: GitHub detected no licence in this paper's repository, YuxuanJiang1/Rock-Token when it was checked. The setup block that called it is withheld with it, because it is written against that body. source_path and origin say where the code lives, to read under the repository's own terms. A run is recorded. |
| decode_id | Ran ran; our contract draft was wrong, not the code |
YuxuanJiang1/Rock-Token/rock_detection/compute_cross_jaccard.py pointer only (licence: NONE) · get_code("20ad231b1e5607ff") How it ranThe body of this sample is withheld: GitHub detected no licence in this paper's repository, YuxuanJiang1/Rock-Token when it was checked. The setup block that called it is withheld with it, because it is written against that body. source_path and origin say where the code lives, to read under the repository's own terms. A run is recorded. |
| fmt | Ran ran; our contract draft was wrong, not the code |
YuxuanJiang1/Rock-Token/rock_detection/compare_checkpoints.py pointer only (licence: NONE) · get_code("2a10cc0c29182e7d") How it ranThe body of this sample is withheld: GitHub detected no licence in this paper's repository, YuxuanJiang1/Rock-Token when it was checked. The setup block that called it is withheld with it, because it is written against that body. source_path and origin say where the code lives, to read under the repository's own terms. A run is recorded. |
| jaccard | Ran ran, honoured the contract we drafted |
YuxuanJiang1/Rock-Token/rock_detection/compute_cross_jaccard.py pointer only (licence: NONE) · get_code("0d6cdf1f7d83f53b") How it ranThe body of this sample is withheld: GitHub detected no licence in this paper's repository, YuxuanJiang1/Rock-Token when it was checked. The setup block that called it is withheld with it, because it is written against that body. source_path and origin say where the code lives, to read under the repository's own terms. A run is recorded. |
| load | Ran ran; our contract draft was wrong, not the code |
YuxuanJiang1/Rock-Token/rock_detection/compare_unrestricted.py pointer only (licence: NONE) · get_code("990a0d73ca1ad6b4") How it ranThe body of this sample is withheld: GitHub detected no licence in this paper's repository, YuxuanJiang1/Rock-Token when it was checked. The setup block that called it is withheld with it, because it is written against that body. source_path and origin say where the code lives, to read under the repository's own terms. A run is recorded. |
| load_rock_ids | Ran ran, honoured the contract we drafted |
YuxuanJiang1/Rock-Token/rock_detection/compute_cross_jaccard.py pointer only (licence: NONE) · get_code("dd218b8873bf5539") How it ranThe body of this sample is withheld: GitHub detected no licence in this paper's repository, YuxuanJiang1/Rock-Token when it was checked. The setup block that called it is withheld with it, because it is written against that body. source_path and origin say where the code lives, to read under the repository's own terms. A run is recorded. |
| load_token_stats | Ran ran; our contract draft was wrong, not the code |
YuxuanJiang1/Rock-Token/rock_detection/funnel_plot.py pointer only (licence: NONE) · get_code("6c4fa6154f037dfc") How it ranThe body of this sample is withheld: GitHub detected no licence in this paper's repository, YuxuanJiang1/Rock-Token when it was checked. The setup block that called it is withheld with it, because it is written against that body. source_path and origin say where the code lives, to read under the repository's own terms. A run is recorded. |
| per_token_kl | Ran ran, honoured the contract we drafted |
YuxuanJiang1/Rock-Token/rock_detection/compare_kl_evolution.py pointer only (licence: NONE) · get_code("ddb9a49095bb09de") How it ranThe body of this sample is withheld: GitHub detected no licence in this paper's repository, YuxuanJiang1/Rock-Token when it was checked. The setup block that called it is withheld with it, because it is written against that body. source_path and origin say where the code lives, to read under the repository's own terms. A run is recorded. |
| per_token_stats | Ran ran, honoured the contract we drafted |
YuxuanJiang1/Rock-Token/rock_detection/compare_checkpoints.py pointer only (licence: NONE) · get_code("b8a19cad03629b57") How it ranThe body of this sample is withheld: GitHub detected no licence in this paper's repository, YuxuanJiang1/Rock-Token when it was checked. The setup block that called it is withheld with it, because it is written against that body. source_path and origin say where the code lives, to read under the repository's own terms. A run is recorded. |
| reverse_kl_logit_grad | Ran ran; our contract draft was wrong, not the code |
YuxuanJiang1/Rock-Token/rock_detection/compute_logit_gradients.py pointer only (licence: NONE) · get_code("14f5c99e3aae37dd") How it ranThe body of this sample is withheld: GitHub detected no licence in this paper's repository, YuxuanJiang1/Rock-Token when it was checked. The setup block that called it is withheld with it, because it is written against that body. source_path and origin say where the code lives, to read under the repository's own terms. A run is recorded. |
Of 11 samples lifted from this paper's linked repositories, 11 carry a function identity and 0 do not. Of the 11 distinct functions with an identity, 0 are also present in other papers whose code we hold (0 with at least one paper dated earlier, 0 the same month, 0 later) and 11 are only here among the 52,977 papers whose code Syntology holds (as of 2026-09-26).
Only here means: among the 52,977 papers whose code Syntology holds (51,887 with lifted samples, 1,090 through their assemblies only), as of 2026-09-26, no other paper's lifted samples or current assembly bricks contain an identical function body after normalisation.
Presence, not direction: an identical function body in two papers' repositories does not say who wrote it, which came from which, or that either paper's method rests on it. Matching is exact after normalisation, so a renamed or edited function is not matched. Dates are when papers appeared, not when code was written. Only .py files the harvester lifted are covered; notebooks are not read. A new-style arXiv id YYMM.NNNN dates a paper to the month 20YY-MM; otherwise Paper.year dates it to the year; otherwise it is undated. Dates are when the paper appeared, not when the code was written. Read 2026-11-22.
Some links come from the archived Papers with Code dataset (CC BY-SA 4.0): attribution and licence.
While recent work in Reinforcement Learning with Verifiable Rewards (RLVR) has shown that a small subset of critical tokens disproportionately drives reasoning gains, an analogous token-level understanding of On-Policy Distillation (OPD) remains largely unexplored. In this work, we investigate high-loss tokens, a token type that-as the most direct signal of student-teacher mismatch under OPD's pertoken KL objective-should progressively diminish as training converges according to existing studies; however, our empirical analysis shows otherwise. Even after OPD training reaches apparent saturation, a substantial subset of tokens continues to exhibit persistently high loss; these tokens, which we term Rock Tokens, can account for up to 18% of the tokens in generated outputs. Our investigation reveals two startling paradoxes. First, despite their high occurrence frequency providing a disproportionately large share of total gradient norms, Rock Tokens themselves remain stagnant throughout training, resisting teacher-driven corrections. Second, through causal intervention, we find that these tokens provide negligible functional contribution to the model's actual reasoning performance. These findings suggest that a vast amount of optimization bandwidth is spent on structural and discourse residuals that the student model cannot or need not internalize. By deconstructing these dynamics, we demonstrate that strategically bypassing these "stumbling blocks" can significantly streamline the alignment process, challenging the necessity of uniform token weighting and offering a more efficient paradigm for large-scale model distillation. Code is available at: https://github.com/YuxuanJiang1/Rock-Token Recent studies in Reinforcement Learning with Verifiable Rewards (RLVR) [15,16,22] have revealed that not all tokens contribute equally to model learning: a small subset of critical tokens, such as high-entropy "forking tokens" at decisive branching points, disproportionately drives reasoning gains [72]. While this token-level perspective has reshaped our understanding of RLVR dynamics, an analogous investigation in OPD remains largely unexplored, despite OPD's reliance on dense token-level supervision. In this work, we initiate such an investigation by focusing on a token type uniquely meaningful to the OPD paradigm: high-loss tokens. Unlike RLVR, where token importance Preprint.
The same record, over MCP at https://syntology.ai/mcp:
get_harvested_code_for_paper("2605.09253")
get_code_for_paper("2605.09253")
have("2605.09253")
The run record, dated, one paper per request, free:
curl https://syntology.ai/api/ran/2605.09253.json
A badge for a README (the split and the date, never a ratio):
[](https://syntology.ai/paper/2605.09253)
Connect an agent — have() is free.