thinker:openalex-A5128257393Max Loeffler
Authored papers (2)
- Anatomy of Post-Training: Using Interpretability to Characterize Data and Shape the Learning Signal2026
Interpretability can audit a preference dataset before optimization and predict what DPO will teach a model: on Dolci, clustering SAE features and testing chosen-minus-rejected activation differences predicted post-DPO feature shifts with R²=0.9 (feature-conditioned) and R²=0.58 (prompt-conditioned). The method is a two-view hypothesis-generation pipeline. A BatchTopK SAE on Llama-3.1-8B supplies the feature clusters (814 response-feature clusters, 512 data clusters over 259,842 pairs), and Welch tests with split-half sign-consistency filtering rank which concepts the labels reward. The pipeline flags Dolci DPO as teaching compliance with unsafe queries, and benchmarks such as HarmBench and XSTest confirm degraded safeguards after DPO. It also flags physics sycophancy, questionable fanfiction, benchmark recognition and hallucinated URLs. Four interventions (data filtering, inoculation prompting, activation steering, reward shaping) are unified as 'explaining away' a concept from an exponential tilt of the policy. Reward shaping is the strongest: at λ=−5 it cuts Llama-3.1-8B harmful-response rate from 21.7% to 4.0% while keeping OLMES accuracy near DPO's 0.434 (0.425), and it also amplifies traits such as playfulness. Interventions fail when the target concept is entangled with others or sparse in the data: style attributes leak off-target, and only URL hallucination improved significantly. The paper argues that post-training should treat the dataset as something to audit and the reward as something to sculpt, and that future protocols need concept hierarchies rather than flat, independent concept axes.
Reasoning models generate chains of thought that are frequently performative rather than causally necessary for reaching the correct answer: on MMLU recall questions, activation probes decode the model's final answer from internal representations far earlier in the chain-of-thought than a CoT language monitor can, demonstrating that the visible reasoning tokens are generated after the belief is already settled. Tested across DeepSeek-R1 671B and GPT-OSS 120B using three complementary methods — activation probing, early forced answering, and a CoT monitor — the analysis reveals a task-difficulty gradient: the gap between probe-decodable certainty and monitor-detectable certainty is large for easy MMLU items but narrows substantially for difficult GPQA-Diamond multihop questions, where genuine step-by-step uncertainty persists longer. The paper introduces probe-guided early exit, which terminates generation once probe confidence passes a threshold, cutting token counts by up to 80% on MMLU and 30% on GPQA-Diamond while preserving accuracy. A key finding that complicates a purely dismissive view is that inflection-point behaviors — backtracking and 'aha' moments — appear almost exclusively in traces where probes register large belief shifts, meaning these surface signals do track genuine uncertainty rather than being uniformly theatrical. The paper argues that this dissociation between internal belief and continued CoT generation is systematic and exploitable, positioning attention probing as both a diagnostic instrument for faithfulness research and a practical mechanism for adaptive computation.
More papers — OpenAlex / S2
Affiliations (2)
- Harvard University(institute)
- Goodfire(institute)
Co-authors (12)
- Atticus Geiger13 shared
- Jack Merullo10 shared
- Owen Lewis10 shared
- Raphaël Sarfati10 shared
- Siddharth Boppana10 shared
- Ekdeep Singh Lubana9 shared
- Leon Bergen9 shared
- Sidharth Baskaran9 shared
- Thomas McGrath9 shared
- Usha Bhalla9 shared
- Daniel Balsam6 shared
- Dhruvil Gala6 shared
Their work is cited by (3)
Other inbound relations (1)
Recent mentions (3)
- papers-typedbergen-2026-anatomy-post.md
- papers-typedboppana-2026-reasoning.md
- papers-typed
boppana-goodfire-reasoning-theater-2026.md